docs(phys L.2g): slice 1 shipped handoff + B.4 gap discovery + plan-of-record

L.2g slice 1 is CODE-COMPLETE: parser + registry mutator + WorldSession
dispatcher + GameWindow subscriber (4 commits: 2459f28, d538915,
536a608, 108e386). Build clean, 6 new tests pass, baseline-stable
across the full suite. Per-commit + final integration code reviews
all approved.

Visual verification deferred: while running the Holtburg-doorway test,
Phase B.4's outbound Use handler turned out to be unwired. The wire
builders (InteractRequests.BuildUse), classes (SelectionState,
WorldPicker), input-action enums, and keybindings all exist — but
GameWindow.OnInputAction has no case for SelectDblLeft, so the click
silently does nothing. The inbound L.2g chain we just landed can't
fire until something sends an outbound Use.

This commit captures the handoff + reframes next-session work:

  * docs/research/2026-05-12-l2g-slice1-shipped-handoff.md (NEW)
    Full evidence: 4 shipped commits, end-to-end code flow, B.4
    discovery explanation, 4 minor + 1 Important review notes
    (the Important one is a test-coverage gap that the B.4b visual
    test will settle automatically), reproducibility recipe,
    next-session pick.

  * CLAUDE.md
    "Currently in Phase L.2" paragraph: L.2g slice 1 code shipped;
    visual test deferred to B.4b. Next-phase-candidates list:
    L.2g slice 1 (now done) replaced with the B.4b candidate
    pointing at the slice scope.

  * docs/plans/2026-04-29-movement-collision-conformance.md
    L.2g section gains a "Current shipped slice (2026-05-12):" table
    listing the 4 commits.

  * docs/plans/2026-05-12-milestones.md
    M1 phase-list updated: L.2g slice 1 (code) shipped; B.4 renamed
    "B.4 / B.4b" with the gap-discovery note + B.4b shape.

  * docs/ISSUES.md
    New issue #57 (HIGH) for the B.4 interaction-handler gap.
    Promoted to Phase B.4b; will close as
    DONE (promoted to Phase B.4b) when B.4b's design spec lands.

  * Memory file project_interaction_pipeline.md (in personal
    memory dir, not in this commit) updated to reflect reality.

Next session: Phase B.4b (~30-50 LOC, 1-2 subagent dispatches,
~30 min). Subscribe SelectDblLeft -> WorldPicker.Pick ->
InteractRequests.BuildUse -> _liveSession.SendGameMessage. Same
Holtburg-doorway visual test verifies both L.2g slice 1 and B.4b
in one pass.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

docs/ISSUES.md

@@ -46,6 +46,65 @@ Copy this block when adding a new issue:
 
 # Active issues
 
+## #57 — B.4 interaction-handler missing: clicking on doors / NPCs / items silently does nothing
+
+**Status:** OPEN
+**Severity:** HIGH (M1 blocker — demo target *"open the inn door, click an NPC, pick up an item"* is fully blocked)
+**Filed:** 2026-05-12
+**Component:** input / interaction / `GameWindow.OnInputAction`
+
+**Description:** Discovered 2026-05-12 while running the L.2g slice 1
+visual test. Phase B.4 (2026-04-28) shipped half of itself: the wire-
+message builders (`InteractRequests.BuildUse` / `BuildUseWithTarget` /
+`BuildPickUp`), the supporting classes (`SelectionState`, `WorldPicker`),
+the `InputAction` enum entries (`SelectLeft`, `SelectDblLeft`,
+`SelectRight`, `SelectDblRight`, `UseSelected`, `SelectionPickUp`, etc.),
+and the default keybindings (LMB → `SelectLeft`, LMB-dblclick →
+`SelectDblLeft`, RMB → `SelectRight`, R → `UseSelected`, F →
+`SelectionPickUp`). What was never shipped: a case for ANY of those
+actions in `GameWindow.OnInputAction`'s switch. The runtime diagnostic
+`[input] SelectLeft Press` fires when you click — confirming the
+dispatcher resolves the chord — but nothing downstream listens, so the
+click silently does nothing. Neither does double-click, R, or F. The
+inbound side (`MoveToObjectReceived`, `StateUpdated` after L.2g slice 1)
+is wired and ready; the block is purely outbound.
+
+**Root cause / status:** B.4 handler integration step was evidently
+dropped or never landed. Memory file
+`memory/project_interaction_pipeline.md` was updated 2026-05-12 to
+reflect this reality (previous text claimed shipped).
+
+**Files:**
+- `src/AcDream.App/Rendering/GameWindow.cs` — `OnInputAction` switch
+  around line 8546+ has no `Select*` cases.
+- `src/AcDream.Core.Net/Messages/InteractRequests.cs` — wire builders
+  exist but have zero callers in `src/`.
+- `src/AcDream.Core/Selection/SelectionState.cs` — class exists, zero
+  production callers.
+- `src/AcDream.App/Rendering/WorldPicker.cs` — class exists, zero
+  production callers.
+- `src/AcDream.UI.Abstractions/Input/KeyBindings.cs:300-320` — bindings
+  for `SelectLeft` / `SelectDblLeft` / `SelectRight` / `SelectDblMid`
+  exist.
+
+**Research:** [docs/research/2026-05-12-l2g-slice1-shipped-handoff.md](research/2026-05-12-l2g-slice1-shipped-handoff.md)
+"Why the visual test is deferred" section has the full investigation.
+
+**Acceptance:** Double-left-clicking on a door in the Holtburg inn
+doorway sends a `0xF7B1 / 0x0036 Use` to the server, the server flips
+the door's `Ethereal` bit and broadcasts `SetState (0xF74B)`, the
+L.2g-slice-1 chain mutates `ShadowObjectRegistry`, the
+`CollisionExemption.ShouldSkip` check honors it, and the player can
+walk through the doorway. Visual verification + log grep (per the
+L.2g handoff's reproducibility recipe) both pass.
+
+**Status promotion:** This is a phase-sized follow-up (estimated
+30-50 LOC, ~30 min). Promoted to **Phase B.4b** in the L.2 milestone
+context and the CLAUDE.md "Next phase candidates" list. Will be closed
+as `DONE (promoted to Phase B.4b)` once that phase's design spec lands.
+
+---
+
 ## #55 — Static-entity slow path reports ~1.45M `meshMissing` per 5s at r4 standstill
 
 **Status:** OPEN