fix(physics): restore retail path-6 collision response

docs/research/2026-07-30-response-layer-edge-family-pseudocode.md

@@ -1,14 +1,17 @@
 # P2 — Collision response-layer edge family: port-ready pseudocode
 
-**Status: RESEARCH PASS COMPLETE (2026-07-30); RETAIL RESPONSE ORDER PORTED
-(2026-07-31).** Originally a research-only doc for Campaign P Slice P2
-(`docs/plans/2026-07-29-physics-parity-campaign.md` §P2); a same-day
-implementation session landed TS-1's retirement and AP-7's fix, attempted
-TS-4 per this doc's own §6 Step 3 fixture-first order, reproduced the
-historical wedge, and stopped — see §7 item 6 for the full capture and
-root-cause diagnosis. TS-4 is NOT retired; its shortcut stays in place.
+**Status: RETAIL RESPONSE ORDER COMPLETE (2026-07-31).** Originally a
+research-only doc for Campaign P Slice P2
+(`docs/plans/2026-07-29-physics-parity-campaign.md` §P2). The first TS-4
+attempt reproduced a steep-roof fixed point because its fixture discarded
+the accepted contact state between frames; §7 item 6 preserves that useful
+failure analysis. Slice 1B then completed the exact nested edge/StepDown
+dispatcher, Slice 2A restored StepDown's mandatory Placement tail, and Slice
+2B repeated the removal with production-shaped state carry and the complete
+direction matrix. TS-4 is now retired; see §10.
 #166 got a reattribution note in ISSUES.md rather than new code (per §3).
-#116 remains untouched (oracle-first, out of implementation scope).
+#116's former skipped wall control is now active with the retail Path-6
+first-frame hard stop and next-frame slide chronology.
 **Headline findings that change the plan's assumptions:** TS-1 was already
 substantially ported (the register row and plan phrasing were stale — see
 §2); the one real gap needed no code change (acdream's unified world-space
@@ -17,17 +20,11 @@ substantially ported (the register row and plan phrasing were stale — see
 NOT about a literal `PhysicsState.Sledding` auto-toggle at all (see §3);
 AP-7's L.3c regression does not reproduce on the production graphical
 root-motion path post-R6, and now ports retail's confirmed 0.25f threshold
-(see §1); TS-4's shortcut removal is coupled to TS-1's completion and
-reproduces a wedge even after TS-1 lands — see the §7 item 6 update for the
-precise mechanism (Phase 3 of `TransitionalInsert` is structurally
-unreachable from Path 6's unconditional `SetCollide`, which returns
-`Adjusted` without repositioning the sphere) and what a future attempt
-needs to check first; #116 remains a genuine oracle-first research item
-needing live cdb/Ghidra, not an implementation item (see §5). Read §6 (port
-order) before starting further implementation on this family. Campaign P
-Slice 1B subsequently performed that fresh `transitional_insert` read and
-removed AP-3, AP-4, AD-53, and AD-54; the exact closeout and controls are in
-§8. TS-4 remains deliberately unchanged.
+(see §1). The apparent TS-4 wedge was a harness-state defect, not a reason to
+retain a product compensation: production-shaped vertical, inward,
+tangential, uphill, and downhill histories now pass graph/flat raw-bit
+parity without the shortcut. Campaign P Slice 1B removed AP-3, AP-4, AD-53,
+and AD-54 (§8); Slice 2A retired AP-5 (§9); Slice 2B retired TS-4 (§10).
 
 Every claim below is tagged **FACT** (grep/read-verified against the
 named-retail decomp, the register, ISSUES.md, or current acdream source
@@ -599,12 +596,13 @@ manufacture a caller that sets it.
 
 ## 4. TS-4 — Path-6 steep-poly shortcut removal
 
-### The current shortcut (FACT, `src/AcDream.Core/Physics/BSPQuery.cs:2149-2266`)
+### The former shortcut (HISTORICAL FACT; removed by Slice 2B)
 
-Path-6 (the default `sphere_intersects_poly → collide_with_pt / SetCollide`
-dispatch) tests each hit polygon's world-space normal. For BOTH sphere0
-(feet) and sphere1 (head), if `worldNormal.Z < PhysicsGlobals.FloorZ`
-(steeper than ~49° from horizontal), acdream takes a SPECIAL BRANCH:
+Before Slice 2B, Path-6 (the default
+`sphere_intersects_poly → collide_with_pt / SetCollide` dispatch) tested each
+hit polygon's world-space normal. For BOTH sphere0 (feet) and sphere1 (head),
+if `worldNormal.Z < PhysicsGlobals.FloorZ` (steeper than ~49° from
+horizontal), acdream took a SPECIAL BRANCH:
 projects the move along the steep face, writes
 `collisions.SetCollisionNormal(worldNormal)` **and**
 `collisions.SetSlidingNormal(worldNormal)`, and returns
@@ -653,7 +651,7 @@ and — when that surface turns out too steep to be walkable —
 `EdgeSlideAfterStepDownFailed` → `CliffSlide`/`PrecipiceSlide` (TS-1's
 domain, §2 above).
 
-### TS-4 port shape (FACT-grounded, mechanically simple)
+### TS-4 port shape (COMPLETED 2026-07-31)
 
 Delete both `if (worldNormal{0,1}.Z < PhysicsGlobals.FloorZ) { ... return
 TransitionState.Slid; }` blocks (`BSPQuery.cs:2200-2215` and
@@ -666,7 +664,7 @@ retires both `SetSlidingNormal` write sites (satisfying DO-NOT-RETRY §0
 item 1 permanently — deleted, not just avoided) with no replacement logic
 needed at this layer.
 
-### ⚠️ Port-order coupling with TS-1 (INFERENCE, but directly evidenced by the shortcut's own commit history)
+### Port-order coupling with TS-1 (historical guard, now satisfied)
 
 **This is the single most important sequencing fact in this whole
 document.** The shortcut's comment proves TS-1's retail-faithful chain
@@ -1182,3 +1180,42 @@ The #273, #271, #185, StepUp, transition-retry, and TS-4 controls remain
 unchanged. AP-5 is retired; TS-4 is intentionally untouched.
 
 ---
+
+## 10. Campaign P Slice 2B closeout — TS-4 retired (2026-07-31)
+
+A fresh read of `BSPTREE::find_collisions` (`0x0053A440`) confirms the exact
+Path-6 split. A primary/foot-sphere polygon hit transforms the polygon normal,
+calls `SPHEREPATH::set_collide`, writes `LandingZ`, and returns `ADJUSTED_TS`
+(`0x0053A7B3..0x0053A7DC`) regardless of steepness. Only when that sphere is
+clear does a secondary/head-sphere hit write `collision_normal` and return
+`COLLIDED_TS` (`0x0053A793..0x0053A7A4`). Neither branch writes
+`sliding_normal`.
+
+Both parsed-graph and prepared-flat Path-6 implementations now follow that
+split exactly. The steepness branches, in-place tangent projection, and BSP-
+layer `SetSlidingNormal` writes are deleted. Exact site tests compare the two
+representations by raw bits and pin every mutated and preserved field:
+
+- foot: `SetCollide`, candidate backup, transformed `StepUpNormal`,
+  `WalkInterp=1`, `WalkableAllowance=LandingZ`, `Adjusted`;
+- head: `CollisionNormal`, `Collided`, with no `SetCollide` state mutation;
+- both: a pre-existing sliding normal is preserved byte-for-byte.
+
+The failed first removal was a test-harness lesson, not a retail exception.
+Its gravity-only replay always passed `isOnGround:false` and discarded each
+accepted result's Contact/OnWalkable bits, making the nested edge/StepDown
+chain impossible to exercise on the next frame. The replacement replay uses
+the same retained `PhysicsBody` chronology as production. Parsed graph and
+prepared flat now match by raw result/body bits for vertical roof descent,
+downhill input, uphill pressure (including a no-launch/no-bounce assertion),
+tangential roof travel, inward-plus-tangent wall travel, and flat-roof ledge
+rejection. Every trace asserts finite bounded motion and signed-plane non-
+penetration. The former #116 D4 control is active: its primary-sphere hit
+hard-stops frame one through SetCollide, then the accepted persistent normal
+permits the downward slide on frame two.
+
+The complete historical #273/#271/#269/#265/#185/#137/#116/cellar/roof
+matrix and the full Core Release suite pass without a replacement
+compensation. TS-4 is retired.
+
+---