cdb attached to retail at a Holtburg cottage door while user walked the
inside-out off-center scenario. Three trace iterations identified that
retail's collision-recording happens via SPHEREPATH::set_neg_poly_hit
(fires hundreds of times during inside-out walk), NOT via the more
obvious-named COLLISIONINFO setters (which fire 0 times). Apparatus
scripts at tools/cdb/door-inside-out-v[1-3].cdb + symbol-probe.cdb.
Our codebase has NegPolyHitDispatch defined but never called. The
downstream TransitionalInsert NegPolyHit handler was a stub. Two-part
fix landed:
1. BSPQuery.FindCollisions Path 5 (Contact branch) restructured —
distinguishes full hit (hit0 == true → StepSphereUp) from near-miss
(hit0 == false but hitPoly0 != null → NegPolyHitDispatch). Mirrors
retail BSPTREE::find_collisions at
acclient_2013_pseudo_c.txt:0053a630-0053a6fb.
2. Transition.TransitionalInsert NegPolyHit handler — dispatches to
step_up + step_up_slide (NegStepUp=true) or records collision
normal + returns Collided (NegStepUp=false). Mirrors retail
CTransition::transitional_insert at
acclient_2013_pseudo_c.txt:0050b7af-0050b7e6.
Tests: all 11 fix-relevant + regression tests pass including issue #98.
VISUAL VERIFICATION (user-driven inside-out off-center): still squeezes
through. Diagnostic [neg-poly-dispatch] probe shows ZERO hits in
production. The Path 5 restructuring doesn't surface NegPolyHit
because our SphereIntersectsPolyInternal only sets hitPoly on FULL
hits — retail's sphere_intersects_poly sets var_5c (closest polygon)
even on near-misses via BSP-traversal side effect.
Remaining fix (next session): add near-miss polygon recording to
SphereIntersectsPolyInternal. Once it sets hitPoly on near-miss BSP
traversal, the Path 5 NegPolyHit dispatch (this commit) will fire
and the TransitionalInsert handler (this commit) will block.
Full handoff with cdb trace table + next-step plan:
docs/research/2026-05-25-door-bug-cdb-retail-trace-findings.md
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
CornerSlide_AlcoveEastToCottageNorth_ShouldBlock test:
- Registers cottage GfxObj 0x01000A2B (contains north exterior walls)
- Registers cell 0xA9B40150 BSP via dat-direct load (alcove walls)
- Places sphere at (132.95, 16.8, 94) inside alcove near east wall
- Walks sphere +Y 50 times at walk speed (0.05 m/tick)
Result: sphere STAYS at (132.95, 16.8) for all 50 ticks with collision
normal cn=(0.71, -0.71, 0) — the average of alcove east wall normal
and cottage north wall normal at their meeting corner. The corner
handling works correctly in the harness.
So production's inside-out walkthrough is NOT a geometric or BSP
collision-detection bug. The geometry exists, the collision detection
fires symmetrically at corners. The discrepancy must be a STATE
difference between harness and production:
- Real walkable polygons with edges (harness uses big quad)
- Real terrain (harness uses Z=-1000 stub)
- Accumulated body state across many prior ticks (harness uses fresh)
- Possibly cell ping-pong between 0x0150 and 0x0029 in production
Cottage GfxObj wall polygons at the doorway area confirmed:
- North exterior wall east of doorway: polys 0x0032, 0x0033
X=[133.5, 136.3], Y=17.10, Z=[94, 97], normal +Y
- North exterior wall west of doorway: polys 0x0030, 0x0031, 0x0034,
0x0035 (X<131.6 various ranges)
- Lintel polys above doorway: 0x0037, 0x0038, 0x003A, 0x003B at Z>96.5
Next-session moves (per handoff):
1. Replay captured tick 2586 (where sphere went from cell 0x0150 to
0x0029 at X=134.022, way past alcove east wall). Inspect engine
behavior at exactly that tick's body state.
2. cdb attach to retail at Holtburg cottage doorway — verify whether
retail also lets sphere walk through at off-center, OR blocks
cleanly. If retail also allows walkthrough, this might be
retail-faithful behavior we should accept.
Updated handoff: docs/research/2026-05-25-door-bug-inside-out-geometry-gap.md
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Followed up the geometry-gap diagnosis with a wider polygon search.
Result: the cottage's north exterior wall east of doorway DOES exist
in cottage GfxObj 0x01000A2B (polys 0x0032, 0x0033) at
world X=[133.5, 136.3], Y=17.10, Z=[94, 97], normal +Y. Symmetric
polys cover the west side and above the doorway lintel.
The wall SHOULD block sphere at X=133.655 (sphere west edge at 133.175
overlaps wall X range; sphere south edge at 17.11 aligns with wall
at Y=17.10).
New hypothesis: the bug is sphere-vs-corner collision at the meeting
point of cell 0x0150's east wall (X=133.5, Y=[16.5, 17.1]) and the
cottage's north exterior wall (X=[133.5, 136.3], Y=17.10). Cell
transit data shows sphere going from X=132.859 entering alcove to
X=134.022 leaving alcove — sphere reached X=134.022 INSIDE cottage
geometry somehow. The sliding along the slab east face (cn=(+1,0,0)
in captured tick 3254) gradually pushes sphere east. Eventually it
shifts past X=133.5 — the corner where alcove east wall meets cottage
north wall. The corner-handling in our BSP collision may incorrectly
let the sphere slide past, or the alcove cell's east wall and cottage
GfxObj's north wall don't compose correctly at the corner.
Diagnostic apparatus extensions:
- HoltburgLandblockStatics_DatInspection: dumps LandBlockInfo for
landblock 0xA9B4. Shows 114 stabs + 12 buildings. The cottage IS
Building[6] with modelId=0x01000A2B (the GfxObj we already loaded).
- Diagnostic_CottagePolys_NearWalkthroughPosition: widened search
reveals the cottage's full north exterior wall geometry.
- HoltburgCottage_CellPortals_DatInspection: extended with cell
PhysicsPolygon world-frame dump (already in prior commit).
Full updated handoff: docs/research/2026-05-25-door-bug-inside-out-geometry-gap.md
Next-session move: add a "sphere walks +Y from inside alcove at
X near 133" test. If harness slides past the corner like production,
investigate BSPQuery's sphere-vs-edge case. If harness blocks at
corner, the bug is elsewhere (cell 0x0150 BSP not queried, or
cottage GfxObj BSP traversal misses the wall poly).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Added diagnostic apparatus that pinpoints the inside-out walkthrough
as a collision-geometry GAP, not a collision-detection bug.
New tests in DoorBugTrajectoryReplayTests:
- InsideOut_Tick3254_WithCottageWalls_ShouldBlock: hypothesis test that
registered cottage GfxObj 0x01000A2B and replayed the captured tick.
Cottage blocked sphere but with cn=(0,0,1) floor-cap normal, not a
wall normal — first signal that cottage geometry near the sphere
isn't a wall.
- Diagnostic_CottagePolys_NearWalkthroughPosition: dumps cottage polys
near sphere XY=(133.655, 17.59) at any Z. Result: ZERO cottage
polygons in that area. The cottage GfxObj has no geometry where the
sphere walks through.
DoorSetupGfxObjInspectionTests.HoltburgCottage_CellPortals_DatInspection
extended to dump cell 0xA9B40150's 4 physics polygons in world frame:
- floor (Z=94), ceiling (Z=96.5), west wall (X=131.6), east wall (X=133.5)
- All walls only span Y=[16.5, 17.1] — the small doorway alcove volume
- North of Y=17.1, no wall
Captured sphere at (133.655, 17.59) is 0.155 m east of cell east wall
AND 0.49 m north of the wall's Y range. No collision geometry exists
at that XY past Y=17.1. The collision representation has a gap that
the visual cottage covers with a wall.
Production capture confirms the diagnosis: cottage GfxObj fires
[bsp-test] 425 times during inside-out walking — visibility IS
correct post-AddAllOutsideCells fix. Door slab fires 245 times. But
the BSP queries find no polygon at (133.655, 17.6+, 94-95.20). The
slab's east face blocks WEST motion (cn=(+1,0,0) as captured), sphere
free to move +Y past it because no wall is there to block.
Three candidates for next-session investigation:
1. Different cottage GfxObj (Holtburg cottages may be multi-piece)
2. Landblock-baked stab static at the cottage exterior wall location
3. Cottage GfxObj's visual polygons wider than physics polygons (dat fact)
Cheapest next step: add LandblockStatics_DatInspection test that
loads LandBlockInfo 0xA9B4FFFE + iterates StaticObjects + prints
every entity at world XY in [131,135] x [16,19]. Reveals what other
entities live at the cottage doorway.
Full handoff: docs/research/2026-05-25-door-bug-inside-out-geometry-gap.md
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The Geometric_DoorSlabZRange_AbovePlayerSphereTop test was computing
slabWorldZBottom as (entity.Z + partFrame.Z) — assuming the slab's
local Z=0 was its bottom. Actually checking the dat shows the slab's
PhysicsPolygons local AABB is min=(-0.954, -0.134, -1.236) max=(0.971,
0.127, 1.255) — the slab's local origin is at its GEOMETRIC CENTER,
not the bottom. With partFrame.Z=1.275 lifting the origin, the slab
world Z is actually [94.139, 96.630], not [95.375, 97.865].
Corrected test now computes both slabLocalZMin and slabLocalZMax from
the polygon vertices and asserts the opposite (correct) geometric fact:
the slab IS at sphere height — overlap from Z=94.139 to Z=95.20 (1.061
m of vertical overlap with the player's sphere). The slab is NOT a
lintel that misses the sphere; it should collide.
Test renamed: Geometric_DoorSlabZRange_AbovePlayerSphereTop →
Geometric_DoorSlabAtSphereHeight_OverlapsInZ.
Handoff doc 2026-05-25-door-bug-partial-fix-shipped.md updated with
the corrected analysis. The "next investigation candidates" list now
points toward cdb attach to retail as the highest-ROI option, since
the BSP collision IS active at sphere height but production still
shows asymmetric walkthrough behavior. The bug is in either the
GetNearbyObjects coverage at primary-cell boundaries, the BSP
polygon partial-overlap handling, or missing cell-BSP collision for
cottage doorway walls.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Built three new tests to investigate the inside-out asymmetric collision
that persists after the AddAllOutsideCells coord fix:
1. Directional_OutsideIn_SouthApproach_BlocksAtSlabSouthFace — sphere
south of door moving NORTH; expects block with cn.Y less than -0.5
2. Directional_InsideOut_NorthApproach_BlocksAtSlabNorthFace — sphere
north of door moving SOUTH; expects block with cn.Y greater than +0.5
3. Geometric_DoorSlabZRange_AbovePlayerSphereTop — pins the slab Z
range vs sphere top math
BOTH directional tests PASS — collision is symmetric at unit-test level.
The asymmetric production bug therefore comes from something the unit
tests do not capture (multi-tick state, cell-tracking flicker, walkable
polygon edge interactions).
The geometric pin test reveals the real story: Setup 0x020019FF places
the part-0 BSP slab 1.275 m ABOVE the entity origin via
PlacementFrames[Default][0].Origin. With the cottage door entity at
world Z=94.1, the slab world Z range is [95.375, 97.865]. Player sphere
top reaches Z=95.20. The slab BOTTOM is 0.175 m ABOVE the sphere top —
the slab NEVER collides with the player.
The slab is a LINTEL (door frame above the doorway), not a leaf. The
door's only effective collider at sphere height is the 0.10 m radius
foot cylinder. The directional tests pass because the cylinder blocks,
not the BSP.
User-reported inside-out off-center walkthrough is the sphere walking
AROUND the foot cylinder (sphere reach 0.48 + cyl 0.10 = 0.58 m; any
sphere center over 0.58 m from cylinder center passes freely). The
visual "body partially intersects door" is the character model
occupying the visual door volume while the collision sphere passes
beside the cylinder.
Reframed handoff in docs/research/2026-05-25-door-bug-partial-fix-shipped.md
points to three candidate next-step investigations:
- Retail-faithfulness audit on setup.Radius / setup.Height interpretation
- Re-inspect door parts 1+2 (GfxObj 0x010044B6) for missed physics shapes
- Test the cottage cell BSP (cell 0x0150 walls) + door together — the
COMBINED collision may be what retail relies on
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
CellTransit.AddAllOutsideCells assumed sphere coords were absolute world
coords (subtracting lbXf = 0xA9 * 192 = 32448 from the sphere position).
Production has used landblock-local coords since Phase A.1
(streaming-center landblock at world origin), so the subtraction
produced localX = -32316, gridX = -1346 → out-of-range → early return
→ ZERO outdoor cells added.
For outdoor primary cells the bug was masked by GetNearbyObjects's
radial sweep. For indoor primary cells (where #98 gates the outdoor
sweep), the door's outdoor cell 0xA9B40029 never reached
portalReachableCells, the door's BSP was never queried, and the player
walked through Holtburg cottage doors unimpeded.
Fix: AddAllOutsideCells treats worldSphereCenter as landblock-local
directly. Matches retail CLandCell::add_all_outside_cells which uses
the per-cell 6-byte landblock-relative position struct.
Existing CellTransitAddAllOutsideCellsTests + CellTransitFindCellSetTests
updated to use landblock-local sphere coords (they were the only callers
using the world-coord convention; production never did).
Apparatus shipped:
- DoorBugTrajectoryReplayTests — live-capture-driven replay harness
that pinpointed the bug per-field at unit-test speed (<500ms iteration)
- AddAllOutsideCells_LandblockLocalSphere_AddsDoorOutdoorCell — direct
unit test that demonstrates the fix
- FindTransitCellsSphere_IndoorExitPortal_AddsOutsideForCapturedSpherePos
— verifies cell-portal traversal at the captured sphere position
- DoorSetupGfxObjInspectionTests.HoltburgCottage_CellPortals_DatInspection
— dat-direct EnvCell + Environment.Cells + portal-poly inspector
- Fixture: tests/AcDream.Core.Tests/Fixtures/door-bug/live-capture.jsonl
(tick 13558 walkthrough + tick 22760 outdoor block)
Visual verification (user-driven at Holtburg cottage door, ~50cm off-center):
- outside→inside RUN: now BLOCKS (was: walks through)
- outside→inside WALK: presumed blocks (not retested)
- inside→outside RUN: PARTIAL — body intersects door, sphere slides through
- inside→outside WALK: same partial behavior
The remaining inside→outside asymmetry is a SEPARATE bug in BSP
collision response for two-sided polygons. The [bsp-test] probe now
fires 245 times for the door entity from indoor (was 0 pre-fix) —
door IS being queried; the BSP polygon-level collision response is
the new bug. Handoff at
docs/research/2026-05-25-door-bug-partial-fix-shipped.md.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the puffed-up framing in the prior task7-shipped handoff.
Honest summary: no user-visible bug fixed this session. Off-center
and inside-out door walk-through still 100% reproducible. The 4
commits shipped real infrastructure (multi-part registration + the
GetNearbyObjects dedup fix that would have silently broken any
future multi-part feature) but no observed behavioral change.
Also explicitly retracts the "step-up is the bug" hypothesis from
the prior handoff doc — ACDREAM_DUMP_STEPUP=1 in the apparatus
produced no stepup: ENTER lines, so DoStepUp wasn't even being
called. That hypothesis was over-reach from an inference I should
not have inflated to a conclusion.
Recommends the apparatus-replay pattern (same one that closed issue
#98 after 6+ speculation rounds): live capture via
ACDREAM_CAPTURE_RESOLVE → harness replay test → first per-field
divergence names the broken assumption.
DO-NOT list for the next session: do not redo the multi-part work,
do not speculate-and-fix, do not relaunch with more probes hoping
for an obvious signal.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Visual verification of Task 7 ship: doors block at dead-center (the
small Cylinder catches) but the BSP slab doesn't catch off-center
or inside-walking-out approaches. Probe-instrumented live capture
proves multi-part registration is correct — every door spawns with
shapes=cyl1+bsp1, and the BSP part is visited 135 times for a single
door at player approaches as close as 0.42 m, with cacheHit=True.
But zero [resolve-bldg] attributions for the BSP shape.
Three artifacts added:
1. TransitionTypes.cs — new [bsp-test] probe in the BSP collision
dispatch, fires BEFORE the cache lookup. Mirrors [cyl-test] on
the Cylinder branch. Distinguishes "cache miss → silent skip"
from "queried but no hit" (the latter doesn't show up in
[resolve-bldg] which only fires on attributed hits).
2. DoorCollisionApparatusTests.cs — new grounded test
(Apparatus_Grounded_50cmOffCenter_*) attempts to reproduce the
production bug via a seeded PhysicsBody (Contact + OnWalkable
+ ContactPlane + WalkablePolygon). Currently doesn't reproduce
because the apparatus's stub-terrain + synthetic-floor setup
diverges from production's real Holtburg geometry. Captured as
"documents-the-bug" — flip the assertion shape when the fix
lands.
3. docs/research/2026-05-24-door-collision-task7-shipped-but-bug-remains.md
— full session handoff. Identifies the remaining bug as a Path 5
(Contact branch + StepSphereUp) misbehavior at thin tall
obstacles, not in the multi-part registration we just shipped.
Leading hypothesis: DoStepUp's downward probe finds the same
flat floor on the OTHER side of the door (Holtburg cottages have
no Z change between exterior and interior floor), declares
step-up success, BSP collision returns OK, sphere walks through.
Recommended next move: relaunch with ACDREAM_DUMP_STEPUP=1 to
verify the hypothesis.
What this commit DOES NOT do: fix the remaining step-up bug. The
A6.P4 multi-part registration foundation is correct and stays.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes the M1.5 "doors don't block in production" bug (alongside the
foundation fix at 3b7dc46). Server-spawned entities (doors, NPCs,
chests, items) now register one ShadowEntry per collision shape —
matching retail's CPhysicsObj-with-CPartArray model
(acclient_2013_pseudo_c.txt:286236) — instead of one Cylinder
approximation per entity.
Before:
RegisterLiveEntityCollision picked ONE shape via a CylSphere → Radius
→ Sphere cascade, registered as a single Cylinder. Doors got a
14 cm × 20 cm cylinder from setup.Radius — far too narrow to span
the doorway gap. Players could walk through closed doors.
After:
- ShadowShapeBuilder.FromSetup emits N shapes:
• one Cylinder per CylSphere
• one Cylinder per Sphere (only when no CylSpheres — retail
convention)
• one BSP shape per Part with a non-null PhysicsBSP
- Caller substitutes the real BoundingSphere.Radius from
PhysicsDataCache for BSP shapes (pure builder's 2.0 placeholder
is tightened to the actual cached value).
- setup.Radius fallback preserved: if the builder produces zero
shapes but Radius > 0, register a Radius-based Cylinder so simple
decorative props don't silently lose collision.
- ShadowObjects.RegisterMultiPart adds N rows, all sharing
entity.Id so the existing UpdatePhysicsState (ETHEREAL flip on
door Use) propagates to every part without changes.
Door 0x020019FF (Holtburg cottage) now registers:
- Cylinder r=0.10 h=0.20 (from the single Sphere)
- BSP from Part 0 = GfxObj 0x010044B5, the 6-face 1.925 m × 0.261 m
× 2.490 m two-sided slab confirmed by
DoorSetupGfxObjInspectionTests
Parts 1 + 2 (GfxObj 0x010044B6, the visual leaves) are visual-only
in the dat by retail design and correctly skipped.
Test impact: 53/53 pass in the shape / registry / door /
cellar-replay scope. App-layer 41/41 pass.
Visual verification needed: launch the client, walk into a closed
Holtburg cottage door from outside (dead center AND ~50 cm
off-center), then walk into it from inside. Door should block all
three approaches. Use the door (click + Use) → door swings open →
walking through passes (ETHEREAL flip via existing SetState path).
Foundation fix dependency:
3b7dc46 fix(phys): GetNearbyObjects dedup-by-entityId silently
drops multi-part shadows
Without 3b7dc46 in place, the BSP shape registered here would be
dropped by GetNearbyObjects's dedup. They land together.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Apparatus test (DoorCollisionApparatusTests) loads door GfxObj 0x010044B5
from the real dat, builds the door entity's shape list via
ShadowShapeBuilder, registers via RegisterMultiPart, and sweeps a player
sphere into the door from three angles. Pre-fix: all three assertions
fail — the sphere walks straight through. The [cyl-test] probe fires
every tick (the small Sphere shape is queried) but no [resolve-bldg] —
the per-Part BSP entry is never reached.
Root cause: ShadowObjectRegistry.GetNearbyObjects deduplicates on
entry.EntityId via HashSet<uint>. Pre-RegisterMultiPart each entity had
exactly one shadow row, so dedup-by-entityId correctly suppressed
multi-cell duplication. After Task 4's RegisterMultiPart introduced
multi-shape rows (1 Sphere + 1 per-Part-BSP for doors; potentially more
for creatures + items), the dedup silently drops everything after the
first. ShadowShapeBuilder emits Sphere shapes before Part-BSPs, so the
Sphere wins and the BSP is dropped — exactly the "Task 7 produced zero
[resolve-bldg] hits" finding from the 2026-05-24 evening handoff.
Fix: dedup on the full ShadowEntry. record-struct equality compares
all fields (EntityId, GfxObjId, Position, Rotation, Radius,
CollisionType, CylHeight, Scale, State, Flags, LocalPosition,
LocalRotation). Distinct shapes of the same entity are not equal and
make it through; the same shape registered in multiple cells (its
fields identical across calls) dedups exactly as before.
Apparatus verification post-fix: all 4 tests pass.
- Dead-center front approach: BLOCKED at Y=11.5 normal=(0,-1,0).
- 50 cm off-center: BLOCKED at Y=11.5 normal=(0,-1,0).
- Back approach from inside: BLOCKED at Y=12.8 normal=(0,+1,0).
- Diagnostic dump: BSP fires at tick 5.
What this fix DOES NOT do: switch live RegisterLiveEntityCollision to
use ShadowShapeBuilder + RegisterMultiPart. That's Task 7 of the
original plan, still reverted. With this foundation fix in place,
Task 7 should now actually deliver door blocking in production.
Test impact: 44/44 in the shape/registry/door scope pass. The broader
Physics suite shows the pre-existing PhysicsResolveCapture
static-state flakiness documented in CLAUDE.md — 6 baseline failures
without my new tests, 10 with them (4 extra are my apparatus tests'
IsPlayer-flag resolves getting captured by a concurrent Capture-test
race). Independent of this fix; verified by isolating each test
class.
Findings + apparatus reasoning:
docs/research/2026-05-24-door-dat-inspection-findings.md
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Read-only deterministic test that opens the real client dat and
dumps Setup 0x020019FF + every GfxObj id it references. Bypasses
PhysicsDataCache's four early-return filters so we see WHAT is in
the dat, not what got into the cache. Skips gracefully when the
dat directory isn't present (keeps CI green).
Result reframes the prior session's investigation:
GfxObj 0x010044B5 (part 0 of the door) DOES have a full door-slab
PhysicsBSP — 6 two-sided (SidesType=Landblock) polygons forming a
1.925m × 0.261m × 2.490m collision volume at frame[0] offset
(-0.006, 0.125, 1.275). Bounding sphere radius 1.975. HasPhysics
flag set. So the handoff's Hypothesis A ("0x010044B5 has no
collision-bearing polygons, only visual") is FALSE.
GfxObj 0x010044B6 (parts 1 + 2, the swinging leaves) IS visual-only
by retail design — HasPhysics clear, PhysicsBSP null, 0 PhysicsPolygons,
but 87 visual Polygons. Our ShadowShapeBuilder skipping these matches
retail's CPhysicsPart::find_obj_collisions short-circuit on
physics_bsp==0 (acclient_2013_pseudo_c.txt:275051) — not a bug.
So the door collision bug is in INTEGRATION, not data. The Task 7
experiment last session registered 0x010044B5's BSP but got zero
[resolve-bldg] attributions. With the data confirmed good, the
next apparatus is a deterministic harness that hydrates 0x010044B5
from a dat dump, registers it via RegisterMultiPart, and sweeps a
player sphere into the door to confirm whether BSP collision fires
in isolation.
Pickup prompt + full reading in
docs/research/2026-05-24-door-dat-inspection-findings.md.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Long session that shipped A6.P4 infrastructure (Tasks 1-6 of the
per-part BSP plan) but discovered the specific shapes we register
from door setup 0x020019FF don't catch the player. Per-part BSP at
0x010044B5 produced ZERO collision attributions in 188K+ resolve
lines despite player walking at doors. Cylinder still blocks
center-only.
Task 7 (refactor RegisterLiveEntityCollision) was implemented and
visually tested, but reverted because the new per-part BSP shape
didn't actually fix the door bug. The infrastructure stays — it's
correctly modeling retail's CPhysicsObj-with-parts model — but the
shapes we feed it need to be re-investigated apparatus-first.
Three hypotheses ranked: (A) part BSPs are visual-only, no collision
polys; (B) building BSP has a wide doorway gap our tiny cylinder
doesn't fill; (C) retail uses Setup.Radius/Height directly. Next-
session move: dump GfxObj 0x010044B5's PhysicsBSP first, then cdb
retail at a doorway.
Recommends: stop speculation, build apparatus, decide fix from
evidence.
#99 stays OPEN. Slice 1's "Closes #99" claim was premature; the real
close requires the per-part BSP work + correct shape identification.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds a one-line diagnostic per Cylinder ShadowEntry tested in
FindObjCollisions, gated on ProbeBuildingEnabled. Useful for the
door-collision investigation surfaced 2026-05-24: tells us whether
the broadphase returned a candidate door AND what CylinderCollision
decided (OK / Collided / Adjusted / Slid).
Off in normal play (probe flag off by default). General-purpose; not
door-specific.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds the regression pin for the _entityShapes cleanup that fca0a13
already implemented (Task 4 folded Task 6's Deregister change in for
the multi-part tests to pass). Verifies that a stray UpdatePosition
after Deregister is a no-op — entity is NOT resurrected via the
_entityShapes rebuild path.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Multi-part entities cached via RegisterMultiPart's _entityShapes now
recompose all part transforms on UpdatePosition (called when the server
broadcasts UpdatePosition (0xF748) for a moving entity). Legacy
single-shape path preserved unchanged for tests + entities that never
went through RegisterMultiPart.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Multi-shape entity registration matching retail's CPhysicsObj model: one
logical entity emits N ShadowEntry rows (one per CylSphere / Sphere /
Part-BSP), all sharing the entity's EntityId. _entityShapes caches the
original shape list per entity for UpdatePosition to recompose part
transforms when the entity moves.
Existing UpdatePhysicsState / Deregister / GetObjectsInCell /
AllEntriesForDebug work unchanged — they iterate by EntityId; multiple
matching entries get handled automatically.
AllEntriesForDebug updated to enumerate all parts per entity (not just
the first) by iterating the first cell that holds entries for each entity.
Single-shape callers that previously relied on deduplicated-by-EntityId
behavior are unaffected since they register exactly one entry per entity.
Six new tests: AllShareEntityId, EmptyShapeList_NoOp,
Deregister_RemovesAllParts, UpdatePhysicsState_PropagatesEtherealToAllParts,
PartsAcrossMultipleCells_AllCellsListed, Register_SingleShapeCompat_Unchanged.
All 24 existing ShadowObjectRegistry tests pass via the unchanged
single-shape Register API. 11/11 CellarUpTrajectoryReplayTests pass.
7/7 ShadowShapeBuilderTests pass.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Local-to-entity transform fields, default-valued so existing single-shape
callers keep working unchanged. RegisterMultiPart (next commit) populates
them per part so UpdatePosition can rebuild the entry's world Position +
Rotation when the entity moves.
All 24 existing ShadowObjectRegistry tests pass (including the 2 new
slice 1 tests from b49ed90).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Pure function translating Setup -> IReadOnlyList<ShadowShape>. Walks
CylSpheres + Spheres (only when no CylSpheres) + Parts (only when the
GfxObj has a non-null PhysicsBSP), using PlacementFrames in the same
Resting -> Default -> first-available priority as SetupMesh.Flatten.
Six tests pin the behavior: door setup produces 4 shapes (0+1+3), sphere
local offset matches Setup data, parts without BSP are skipped, creature
setups with CylSpheres skip Spheres, scale factor multiplies all radii
and offsets, empty setup returns empty list, null setup throws.
No callers in this commit; RegisterMultiPart + the GameWindow callers
follow.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Standalone record representing one collision-bearing shape attached to
a logical PhysicsObj. Foundation for the per-part BSP collision fix
that closes the M1.5 "doors don't block" bug. Spec at
docs/superpowers/specs/2026-05-24-door-collision-per-part-bsp-design.md.
No callers in this commit; integration follows in subsequent commits.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Captures the brainstorm session 2026-05-24 evening after A6.P4 slice 1
(b49ed90) shipped without closing #99. Investigation surfaced the actual
root cause: doors register as a single 14cm × 20cm bounding-cylinder
approximation derived from Setup.Radius/Height fallback. Their real
collision-bearing geometry lives in per-part GfxObj BSPs (3 parts for
Setup 0x020019FF), including the threshold polygon spanning the doorway.
Retail-faithful design: every server-spawned entity registers N shadow
entries (one per CylSphere + one per Sphere + one per Part-with-BSP),
all sharing the same EntityId. UpdatePhysicsState propagates ETHEREAL
flips to all entries via the existing EntityId-iteration path. Unifies
the live-entity and landblock-static registration code paths under one
ShadowShapeBuilder.
Retail anchor: CObjCell::find_obj_collisions → CPhysicsObj::FindObjCollisions
→ CPartArray::FindObjCollisions → CPhysicsPart::find_obj_collisions →
CGfxObj::find_obj_collisions. One PhysicsObj per entity, parts iterated
internally for collision (acclient_2013_pseudo_c.txt:276776-275055).
Five-commit migration sequence; tests at three layers (builder unit tests,
registry behavior tests, live-capture regression pin). Approach A approved
by user 2026-05-24.
Spec stands on its own as M1.5 work; not formally assigned a phase letter
per CLAUDE.md's "don't invent phase numbers on the fly" rule.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes#99 (run-through doors regression from b3ce505).
The b3ce505 stopgap for #98 gates the outdoor 24m radial sweep on indoor
primary cells. Combined with ShadowObjectRegistry.GetNearbyObjects'
"skip outdoor ids" filter on the cellScope-pass loop, this meant doors
registered at outdoor cells (default cellScope=0u for server-spawned
entities at GameWindow.cs:3139) were invisible to spheres on the indoor
side of a doorway threshold — walk-through.
Pre-flight reads found that CellTransit.FindCellSet already adds
outdoor cells to its candidate set when the sphere straddles an
OtherCellId=0xFFFF exit portal (via AddAllOutsideCells triggered by
exitOutside=true inside the indoor-seed BFS). The fix is to stop
filtering those outdoor ids out before iterating, and rename the param
to portalReachableCells to reflect what the set actually contains.
- Q1: Indoor EnvCell.VisibleCellIds is indoor-only in all 16 cottage
fixtures (low 16 bits ≥ 0x0100). OtherCellId=0xFFFF on portals
marks "exit to outdoor world" without naming a specific cellId; the
specific outdoor cell is computed by AddAllOutsideCells from world
XY when the sphere straddles the exit portal.
- Q2: GameWindow.cs:3139 ShadowObjects.Register for server-spawned
entities passes no cellScope → default 0u → outdoor 24m grid
registration. UpdatePosition (line 145) does the same on movement.
Doors are confirmed outdoor-registered.
Slice 1 makes a smaller change than the spec proposed (no new
parameter; just drop the existing filter), because FindCellSet's
existing exit-portal logic already exposes the needed outdoor cells.
The retail-faithful registration-side BuildShadowCellSet refactor and
the b3ce505 gate removal stay scheduled for slices 2-3.
Verification:
- 24/24 ShadowObjectRegistryTests pass (incl. two new slice 1 tests:
IndoorPrimary_OutdoorCellInPortalSet_DoorReturned closes#99;
IndoorPrimary_IndoorOnlyPortalSet_OutdoorRadialStillSkipped
regression-pins #98)
- 11/11 CellarUpTrajectoryReplayTests pass (LiveCompare_FirstCap_
FixClosesCottageFloorCap stays green)
- dotnet build AcDream.slnx: 0 errors, 0 warnings
- Pre-existing 6-8 static-state-leakage failures in serial physics
suite verified unchanged by stash+retest baseline check
Visual verification pending: walk Holtburg cottage doorway from both
sides; door blocks both directions; cellar still climbable.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Self-contained pickup doc for the next session. Combines:
- State summary (what's done, what's open, where we are in M1.5)
- Direction (Option B chosen 2026-05-24 — A6.P4 full then #100)
- Slice 1 pre-flight (Q1 + Q2 to resolve before coding)
- Slice 1 / 2 / 3 implementation plans with commit shapes
- #100 follow-up plan
- Decomp anchors reference card (8 line citations)
- Apparatus inventory (don't rebuild what's already there)
- CLAUDE.md rules that apply
- Copy-paste pickup prompt at the bottom
Cross-references all the canonical artifacts from this saga:
- docs/superpowers/specs/2026-05-24-phase-a6-p4-retail-shadow-architecture.md
- docs/research/2026-05-23-a6-p3-issue98-comparison-harness-findings.md
- docs/ISSUES.md (#98 DONE, #99 OPEN, #100 OPEN)
- memory: feedback_retail_per_cell_shadow_list.md,
feedback_apparatus_for_physics_bugs.md
- commits b3ce505 + b55ae83 (don't redo)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Knowledge-preservation pass after the issue #98 cellar-up fix shipped
(`b3ce505`). Closes the saga's documentation loop and plans the next
phase.
Changes:
- docs/research/2026-05-23-a6-p3-issue98-comparison-harness-findings.md
Appended "Resolution 2026-05-24" section: v3 hypothesis falsified,
actual mechanism (head-bump cottage GfxObj floor poly from below)
confirmed, b3ce505 fix shipped, known door regression flagged.
Memory artifacts cross-referenced.
- docs/ISSUES.md
#98 moved to DONE with full resolution writeup + decomp anchors.
#99 filed: door regression at building thresholds (caused by
b3ce505's indoor-primary gate). Closes via A6.P4.
#100 filed: transparent rectangular patches around houses
(terrain rendering). Bisect found commit 35b37df introduced the
hiddenTerrainCells mechanism that collapses 24m outdoor cells
when buildings sit in them; cottage building only fills part of
its cell so the rest of the 24m cell shows the sky-bleeding gap.
Three fix-path options documented.
- docs/superpowers/specs/2026-05-24-phase-a6-p4-retail-shadow-architecture.md
Full A6.P4 design doc. Three-slice plan: (1) query-side portal
expansion to close#99 while preserving #98 fix, (2) port retail's
BuildShadowCellSet at registration time so per-cell semantics match
`CObjCell::find_cell_list`, (3) remove b3ce505 stopgap entirely.
Decomp anchors, file-by-file plan, risk inventory, open questions.
Memory entries written separately (out-of-tree at
~/.claude/projects/.../memory/):
- feedback_retail_per_cell_shadow_list.md
The architectural lesson: retail uses per-cell shadow_object_list
with portal-aware registration; our landblock-wide spatial
registry diverges at indoor/outdoor seams.
- feedback_apparatus_for_physics_bugs.md
The apparatus-first pattern that cracked the saga: live capture +
fixture dump + replay harness. Template for future physics bugs.
Quote rule: "when a physics bug is resisting and you catch
yourself about to ship 'fix attempt N+1 with no new evidence,'
STOP. Build the apparatus first."
- MEMORY.md index updated with both new entries.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The cellar-up cap was caused by ShadowObjectRegistry.GetNearbyObjects
running its outdoor 24m-grid radial query unconditionally — including
when the moving sphere's primary cell is indoor. The landblock-baked
cottage GfxObj 0x01000A2B (registered with cellScope=0u, i.e.
landblock-wide) was returned for a sphere inside the cellar EnvCell,
and its downward-facing cottage-floor poly at world Z=94 head-bumped
the sphere from below, capping ascent at foot Z=92.74.
Diagnosis this session via the live capture in
a6-issue98-resolve-capture-2.jsonl (92K records, 132 cap events all
with body on the ramp polygon) FALSIFIED the prior "stale ramp
contact plane" hypothesis: the contact plane is correctly the ramp's
plane because the sphere IS on the ramp at the cap. The cap is a
proximate consequence of the cottage GfxObj being queried at all from
an indoor primary cell.
Retail decomp anchor (acclient_2013_pseudo_c.txt):
- 308751-308769: CObjCell::find_cell_list branches on the moving
object's m_position.objcell_id — INDOOR adds only that cell +
portal-visible neighbors via CELLARRAY::add_cell; OUTDOOR adds
all overlapping outdoor cells via CLandCell::add_all_outside_cells.
Object-position-driven, not sphere-radius-driven.
- 309560: CEnvCell::find_collisions calls find_env_collisions
(own cell BSP only) THEN CObjCell::find_obj_collisions on `this`.
- 308916: CObjCell::find_obj_collisions iterates this->shadow_object_list
— strictly per-cell, never landblock-wide.
Combined: a landblock-baked static like the cottage building is added
to outdoor cells' shadow_object_list only (its m_position resolves to
an outdoor cell). An indoor EnvCell's shadow_object_list never
contains the cottage. CEnvCell::find_collisions therefore never tests
the sphere against the cottage. Retail-faithful behavior.
Falsification spike (this session): scoping the cottage to a single
distant outdoor cell instead of landblock-wide caused the harness
LiveCompare_FirstCap test to stop reproducing the cn=(0,0,-1) cap,
confirming the cap is caused by the radial sweep returning the
cottage to an indoor primary.
The fix:
- Add optional `primaryCellId` parameter to
ShadowObjectRegistry.GetNearbyObjects. When indoor (>= 0x0100),
skip the outdoor radial sweep entirely after the indoorCellIds
branch runs. Default 0u preserves prior behavior for
cell-unaware callers (existing tests pass unchanged).
- Transition.FindObjCollisions passes sp.CheckCellId.
- Harness LiveCompare_FirstCap_* flipped to documents-the-fix form
(asserts the downward-facing cottage-floor cap does NOT fire).
Deletes the residual-X-motion test that documented a post-cap
edge-slide — irrelevant once the cap is gone.
This same gate should close the other "Finding 3 family" indoor/outdoor
collision bugs (#97 phantom collisions, indoor sling-out). Visual
verification by the user is the remaining acceptance check before
closing #98.
Verification:
- 11/11 CellarUpTrajectoryReplayTests pass in isolation
- 55 ShadowObjectRegistry + TransitionTypes + PhysicsEngine
+ CellPhysics + CellTransit tests pass
- 8 pre-existing static-state-leakage failures in serial physics
suite are unchanged (verified by stash + retest on baseline)
- dotnet build clean, 0 warnings
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Today's evening session ran from "harness still doesn't reproduce the
cap" → "harness reproduces it" → "wait, the cap is only a symptom, the
real cause is upstream Z drift from the contact plane never refreshing."
The breakthrough question, from the user: "we know how retail OPENs it
from above, how hard can it be to know how to open it from below?" —
which reframed the investigation away from cap-event mechanics (where
six prior attempts looked) and toward "what about our STATE is wrong
when the player is in the cellar but not on the ramp?"
The math: player at cap is 10 m away from the cellar ramp in cell-local
X, but body.ContactPlane is still the ramp's slope plane. AdjustOffset
projects forward motion along that stale slope every tick, lifting Z
by +0.201 m per tick. After enough ticks of horizontal walking, the
head sphere reaches Z=94 and bumps the cottage floor. If the contact
plane refreshed to the flat cellar floor when the player walked off
the ramp, the drift would be zero, the cap would never be reachable.
Next session's task (per the pickup prompt at the bottom of the
findings doc): (1) verify the hypothesis chronologically against the
live capture, (2) find the walkable-refresh gap in
Transition.FindEnvCollisions / SpherePath.SetWalkable, (3) cross-ref
retail's CObjCell::find_env_collisions for the per-tick contact-plane
refresh logic.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The findings doc gets an evening-v2 follow-on documenting:
- GfxObj dump infrastructure shipped (cc3afbc)
- Harness reproduces cap-event collision normal (97fec19)
- Residual +0.0266m X-motion divergence — the new investigation target
- Pre-existing test suite flakiness (out of scope, tracked separately)
CLAUDE.md's "Current A6 phase" block points at the residual divergence
as the next concrete move with the test that gives <1s feedback per fix
attempt.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Apparatus convergence. With the cottage GfxObj 0x01000A2B registered as
a ShadowEntry in BuildEngineWithCellarFixtures, the harness now reproduces
the live cap-event collision normal (cn=(0,0,-1)) exactly, ending the
"harness doesn't reproduce" divergence the prior session's findings doc
identified.
Concretely:
* Adds a minimum-stub landblock (TerrainSurface at z=-1000) so
TryGetLandblockContext succeeds at the cellar XY — production's
FindObjCollisions early-returns without a landblock and would skip
the cottage shadow query.
* Adds RegisterCottageGfxObj that loads the 74-polygon cottage fixture
via GfxObjDumpSerializer.Hydrate, then registers it at the cottage's
world transform (translation (130.5, 11.5, 94.0) + 180° around Z,
derived from the cellar cell's WorldTransform), matching
GameWindow.cs:5893's landblock-baked-static registration shape.
* LiveCompare_FirstCap_HarnessMissesCottageFloorBecauseCottageGfxObjNotRegistered
flips: the cap-normal reproduction is now enforced by
LiveCompare_FirstCap_HarnessReproducesCottageFloorCapNormal.
* The full per-field round-trip uncovered ONE residual divergence:
live preserves +0.0266m of +X motion through the cap event (edge-
slide along the floor in XY); harness blocks ALL motion at the cap.
Captured by LiveCompare_FirstCap_ResidualXMotionDivergence_Docs...
in documents-the-bug form so the next session has a concrete next
target.
Fixture: tests/AcDream.Core.Tests/Fixtures/issue98/0x01000A2B.gfxobj.json
(74 polygons, 6 downward-facing cottage-floor triangles at object-local
Z=0, BSP radius 13.989m matching the live [resolve-bldg] bspR=13.99).
Captured via launch-a6-issue98-cottage-gfxobj-dump.ps1.
In-isolation: all 12 CellarUpTrajectoryReplayTests + 4 GfxObjDumpRoundTripTests
+ 1 new PhysicsDiagnosticsTests pass.
Note on full-suite baseline: the full xUnit serial run shows 8–19
failures depending on order (pre-existing test interaction with shared
statics across PlayerMovementControllerTests, MotionInterpreterTests,
PositionManagerTests, etc.). The flakiness is independent of this
change — confirmed by stashing the harness changes and observing the
same flaky range. Investigating the static-state isolation problem is
out of scope for issue #98; tracked as a follow-up.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Mirror the existing ACDREAM_DUMP_CELLS pattern for GfxObj-owned geometry:
when ACDREAM_DUMP_GFXOBJS lists a hex GfxObj id, the first
PhysicsDataCache.CacheGfxObj for that id writes the full resolved
polygon table to a JSON fixture under
tests/AcDream.Core.Tests/Fixtures/issue98/0x{id:X8}.gfxobj.json (override
dir via ACDREAM_DUMP_GFXOBJS_DIR).
Motivation: the existing [resolve-bldg] probe captures GfxObj-level
metadata (id, BSP root radius, entity origin) but emits
"hitPoly: n/a (BSP path — side-channel not written)" because the
BSPQuery wire site that would populate LastBspHitPoly never landed.
A polygon-level dump at cache time bypasses that gap — one capture run
yields the FULL polygon table, fixture-loadable by the harness's
RegisterCottageGfxObj helper (next commit).
See docs/research/2026-05-23-a6-p3-issue98-comparison-harness-findings.md
for the cottage GfxObj 0x01000A2B context: landblock-baked static at
entity origin (130.5, 11.5, 94.0), responsible for the head-sphere cap
from below at world Z=94.0 that issue #98 is documenting.
Test baseline: 1183 + 8 pre-existing failures (serial run; +5 new tests
all pass; was 1178 + 8 pre-session).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds the "Evening extension v2" paragraph documenting the apparatus
ship + root cause identification from the 2026-05-23 PM-late session.
The block names:
- The four commits that landed (fb5fba6 → 44614ab → 0f2db62 → f29c9d5)
- The apparatus: PhysicsResolveCapture + LiveCompare_* tests
- The root cause: cottage GfxObj 0xA9B47900 (landblock-baked static
building) blocks the head sphere at world Z=94.0 with cn=(0,0,-1)
- User's confirming observation about jumping (rules out step-up
hypotheses)
- The cap geometry math (foot Z=92.74 + sphereHeight 1.20)
- Documents-the-bug pattern for the first-cap test
- Test baseline (1178 + 8 serial)
- Pointer to the new findings doc as canonical pickup
- Concrete next-session move (extract cottage GfxObj polygons)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Session-end documentation for the 2026-05-23 evening session in
which:
1. The PhysicsResolveCapture apparatus shipped (committed earlier
in fb5fba6).
2. A live capture (41K records) drove the first LiveCompare_* tests
in CellarUpTrajectoryReplayTests, two of which PASS bit-perfect.
3. The failing third test pinpointed the cap-event divergence.
4. A second capture (70K records + 16 cell dumps + per-poly probes)
identified the cottage GfxObj 0xA9B47900 as the blocker — a
landblock-baked static building whose floor polygons live in the
GfxObj's BSP, NOT in any cottage cell.
The findings doc has:
- TL;DR + chronological commits
- Apparatus inventory (PhysicsResolveCapture, comparison tests,
fixtures, launch scripts)
- The math: head sphere top at Z=foot+1.68 reaches the cottage floor
at Z=94.0 when foot Z=92.74, matching the observed cap.
- User's confirming observation (cap fires on pure-vertical jump too,
ruling out every step-up / AdjustOffset hypothesis)
- What's NOT yet known (why retail doesn't have this cap; full
cottage GfxObj polygon list)
- Next-session pickup with two ranked options
Adds:
- docs/research/2026-05-23-a6-p3-issue98-comparison-harness-findings.md
- launch-a6-issue98-capture.ps1 (capture-only launch)
- launch-a6-issue98-polydump.ps1 (capture + diagnostic probes + 16-cell dump)
- 13 new cell-dump fixtures (0xA9B40140-0xA9B40142, 0xA9B40144,
0xA9B40145, 0xA9B40148-0xA9B4014F) at 272 KB total. The harness now
has the full 0xA9B4014X neighborhood available for any future
comparison test that needs adjacent cell geometry.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The previous version of LiveCompare_FirstCap_HeadHitsCottageFloor
asserted the harness matched the live cap by per-field diff, which
correctly FAILED with a clear divergence message. Converted it to the
documents-the-bug pattern matching the existing
Harness_Finding_SphereGoesAirborneAtTick1 style: passes WHILE the
harness lacks the cottage GfxObj, and will start failing when the
cottage GfxObj is added — at which point the test should be flipped
to AssertCallMatchesCapture(engine, captured).
Test name now reads as a finding:
LiveCompare_FirstCap_HarnessMissesCottageFloorBecauseCottageGfxObjNotRegistered
Second-capture poly-dump finding (committed in the test's xmldoc): the
live cap event attributes the blocking entity as obj=0xA9B47900 — a
landblock-baked static building (the cottage GfxObj). The cottage's
floor lives in this GfxObj's polygon table as a ShadowEntry, NOT in
any of the cottage's cells. The harness's BuildEngineWithCellarFixtures
intentionally skips RegisterStairRampGfxObj today, so the cottage
floor (downward-facing polygon at world Z=94.0) isn't present — and
the harness doesn't reproduce the cn=(0,0,-1) cap.
Next-session move: extract the cottage GfxObj's full polygon list
from a focused live capture (set ACDREAM_PROBE_BUILDING=1 so the
[resolve-bldg] probe fires per-polygon during the cap), add it to
RegisterStairRampGfxObj (rename to RegisterCottageGfxObj), uncomment
the registration call. The harness should then reproduce live's
cn=(0,0,-1) — at which point the documents-the-bug test starts
failing and should be flipped to the assertion form.
Test baseline maintained: 1178 + 8 pre-existing failures (was
1172 + 8 pre-changes; added 6 tests, all pass under serial run).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The capture apparatus pays off on the FIRST iteration. Three records
sampled from a live cottage-cellar session — tick 0 (spawn at Z=92.53),
tick 376 (player on the cellar ramp at Z=91.49), and tick 1183 (first
cap event, foot Z=92.74 with cn=(0,0,-1)) — replayed against the
harness engine reveal:
- LiveCompare_Tick0_Spawn: PASSES (full round-trip).
- LiveCompare_Tick376_OnRamp: PASSES (ramp walkable
polygon hydrates correctly,
engine reproduces live).
- LiveCompare_FirstCap_HeadHitsCottageFloor: FAILS by exactly the
divergence shape that names
the missing fixture.
The cap-record divergence:
Result.Position: live=(141.3865,7.2243,92.7390)
harness=(141.3599,7.2243,92.7390) (Y slid; X stuck)
Result.CollisionNormal:live=(0,0,-1) ← downward = cottage floor from below
harness=(0,0,+1) ← upward = some other floor
Plus the LiveCompare_FirstCap_DiagnosticDump test (always passes; it's
a probe-firing scratch test) prints every cell polygon in world frame:
Cellar 0xA9B40147 — ceiling polys at world Z=93.80 cover X=133-142,
Y=-1.0-11.5 but NOT the sphere XY of (141.39, 7.03) — at the right
edge of Y=7.03 the ceiling quads are at Y<3.90 or Y>8.70.
Cottage 0xA9B40143 — floor polys at world Z=94.0 cover X=136.7-140.5,
Y=3.9-13.1 but NOT (141.39, 7.03) either — at X=141.39 we are 0.89m
east of the floor quad's rightmost vertex.
Cottage 0xA9B40146 — only 4 walls, no floor.
So both cells we have CAN'T produce the live's cn=(0,0,-1). The actual
blocking polygon must be in a cell or static object we haven't loaded
into the harness yet. The cellar is rectangularly bounded; the cottage
above has a floor that spans the cottage, but the floor polygon RIGHT
ABOVE the ramp top (which is where the freeze fires) is in some OTHER
cell — either a separate cottage-floor sub-cell or a building static
GfxObj.
This is the first evidence-driven step in the saga. Six sessions of
speculation produced ten failed fix shapes; the apparatus produced
this finding in one round trip. Next step: re-capture with
ACDREAM_PROBE_POLY_DUMP + ACDREAM_DUMP_CELLS covering 0xA9B40140-
0xA9B4014F to identify the missing fixture cell.
Adds:
- LiveCompare_Tick0_Spawn / Tick376_OnRamp / FirstCap_HeadHitsCottageFloor
- LiveCompare_FirstCap_DiagnosticDump (always passes; dumps cell polys
in world frame + enables every relevant probe so the captured stdout
shows the harness BSP query path)
- tests/AcDream.Core.Tests/Fixtures/issue98/live-capture.jsonl
(3 representative records from the 41,228-record live capture)
- AssertCallMatchesCapture helper: per-field diff with Vector3 / float
tolerances, reports every divergence not just the first.
Test baseline maintained at 1172 + 8 baseline + 5 new tests pass +
1 known-failing test that pinpoints the bug = 1178 + 9 (where the
new failure is the desired evidence-driven test).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Apparatus only — no fix attempt. Per the systematic-debugging skill's
"3+ failures = question architecture" rule, the 6 hypotheses we
tested speculatively on the harness's airborne-at-tick-1 bug all
failed because we kept guessing what state the harness lacks. This
commit ships the evidence-driven path: capture the EXACT player
ResolveWithTransition call (every input + body-before + body-after +
result) into a JSON Lines fixture, then a comparison test loads the
fixture and replays it against the test engine. The first per-field
divergence pinpoints the missing apparatus state — no more guessing.
Adds:
- src/AcDream.Core/Physics/PhysicsResolveCapture.cs — new static module
with CapturePath (env var ACDREAM_CAPTURE_RESOLVE), PhysicsBodySnapshot
record, JSON Lines writer (thread-safe, flushes per record), process-
exit hook for clean shutdown.
- PhysicsEngine.ResolveWithTransition probe wiring: snapshot body at
method entry, snapshot again before return, refactor the two returns
into one path so the capture call site is single. Filtered to
IsPlayer mover flag so NPC/remote DR calls don't pollute.
- CellarUpTrajectoryReplayTests.cs:
• Capture_WritesJsonLinesRecordsWhenIsPlayerAndEnabled — drives 3
ticks with capture on, reads file back, verifies round-trip of
inputs + body-before/after snapshots.
• Capture_SkipsNonPlayerCalls — drives 3 NPC-style ticks (no
IsPlayer flag), confirms the file is not created.
Off by default. Set ACDREAM_CAPTURE_RESOLVE=<path> to a writable file
path; capture starts on the next player ResolveWithTransition call.
Test baseline: 1172 + 8 pre-existing failures + 2 new smoke tests
that pass = 1174 + 8. Verified by stashed-baseline comparison.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds the canonical pickup document
docs/research/2026-05-23-a6-p3-issue98-harness-handoff.md with:
- TL;DR + session arc (10 commits chronological)
- What the trajectory replay harness IS (committed apparatus)
- Bug 1 status: #98 cellar-up freeze (unfixed, 6 fix shapes failed)
- Bug 2 status: airborne-at-tick-1 (new, 6 hypotheses tested, root
cause not isolated)
- Exclusion list: DO NOT retry any of the 6+6 dead ends
- Apparatus inventory: probes, tests, fixtures, cdb captures
- Recommended next move: side-by-side comparison harness against
live PlayerMovementController state (evidence-first instead of
speculation-first)
- Alternative moves: pivot to other M1.5 issues or M2 prep
- Self-contained pickup prompt at the bottom of the handoff doc
Updates CLAUDE.md's "Current A6 phase" block to point at the new
handoff doc as the canonical resume artifact.
Updates ISSUES.md's #98 entry with the late-day extension findings,
the 6-hypothesis exclusion list, and a pointer to the handoff doc.
Test baseline maintained at 1172 + 8 pre-existing failures.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Multi-step investigation of the airborne-at-tick-1 bug per the
systematic-debugging skill. Several hypotheses tested via the
harness, each producing the same (0,1,0) hit normal at tick 1:
1. WalkablePolygon seeding ADDED to BuildInitialBody (was missing).
PhysicsEngine.cs:665-673 requires body.WalkablePolygonValid +
WalkableVertices to call SpherePath.SetWalkable. With seeded
walkable poly: walkPoly=True survives tick 1 (was False before).
BUT engine still reports hit=(0,1,0) and body goes airborne.
2. Initial Z lift removed (back to 0): same airborne behavior.
3. Synthetic stair GfxObj DISABLED: same (0,1,0) hit. Hit is not
from FindObjCollisions.
4. Stub landblock REMOVED: same (0,1,0). FindObjCollisions early-
returns without landblock context, FindEnvCollisions's outdoor
terrain returns null. Hit is not from terrain.
5. SYNTHETIC BSP attached to cell fixtures (Hydrate sets BSP=null
per its xmldoc; without BSP the indoor branch is skipped, falls
through to outdoor terrain). One-leaf BSP referencing every poly
in cell.Resolved. Indoor BSP path now runs. Same (0,1,0) hit.
Trace timeline at tick 1:
find-start: walkPoly=True, CP valid, oi=0x303 (Contact+OnWalkable)
after-adjust: req=(0,-0.1,0) adj=(0,-0.1,0) — no projection change
before-insert: check=(141.5, 9.4, 91.43)
stepdown-enter (Contact-recovery): stepDown=True, height=0.04
stepdown-after-offset: check=(141.5, 9.4, 91.39) — moved DOWN 0.04
stepdown-after-insert: state=OK, cp=n/a (no walkable found)
stepdown-reject
(second stepdown attempt — same outcome)
after-insert: state=Collided, hit=n/a, walkPoly=False
after-validate: state=OK, hit=(0,1,0), slide=(0,1,0)
oi=0x300 (Contact+OnWalkable CLEARED)
The (0,1,0) hit is set by ValidateTransition between after-insert
and after-validate. ValidateTransition's default-push-up code path
sets UnitZ=(0,0,1), NOT UnitY=(0,1,0). So something INSIDE
TransitionalInsert sets ci.CollisionNormal=(0,1,0) before
ValidateTransition runs (12 SetCollisionNormal call sites in
TransitionTypes.cs — root cause not isolated to one).
Per systematic-debugging skill: 5+ hypotheses tested without
convergence = "question architecture". The bug is hidden deeper
than a single misconfigured init field.
Next session pickup: build a side-by-side instrumentation harness
that mimics PlayerMovementController's EXACT call sequence
(PhysicsBody field state, ResolveWithTransition args, frame
ordering) and compare per-tick divergence against a live capture.
The harness is missing some piece of state production carries
across ticks — find what piece.
Apparatus progress (committed):
- Harness with synthetic stair GfxObj registration (Issue #98 ramp polygon now constructable programmatically)
- Synthetic cell-BSP attachment (AttachSyntheticBsp) — unlocks indoor
BSP collision path for hydrated cell fixtures
- WalkablePolygon seeding in BuildInitialBody (PhysicsBody seeding pattern documented)
- Three diagnostic dump tests for tick-by-tick traces
Test baseline: 1167 + 5 (harness) = 1172 + 8 pre-existing failures.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds two diagnostic-only tests:
- Harness_DiagnosticDump_FirstTenTicks: prints trajectory + resolve
probe lines for the seeded-body path
- Harness_DiagnosticDump_NoBodySeed: same but with body=null, isolating
whether the CP seed contributes to the airborne-at-tick-1 issue
Also adjusts InitialSphereWorld to lift the sphere by 0.05m above
cellar floor (sphere bottom at Z=91.00, not Z=90.95). The lift
should give the engine a clean step-down on tick 1 instead of an
exact-boundary contact.
Experimental finding: NEITHER the no-body-seed path NOR the 0.05m
lift changes the airborne-at-tick-1 behavior. With sphere center
at world Z=91.48 (0.05m + radius above cellar floor at 90.95):
- Tick 1: in=(141.5, 9.5, 91.48), out=(141.5, 9.5, 91.48) — Y move
rejected. hit=yes n=(0,0,1) walkable=False.
- Tick 2+: Y advances by 0.1/tick, Z stays put, onGround stays False.
The hit normal (0,0,1) at tick 1 means the engine treats the cellar
floor polygon as a NON-WALKABLE collision target when the sphere is
seeded grounded above it. The walkability classifier returns False
even though Normal.Z=1.0 > FloorZ=0.6642. This is a real engine bug
worth investigating in a future session — independent of the cellar-up
freeze.
The synthetic ramp polygon registered via RegisterStairRampGfxObj is
NOT reached because the sphere is now airborne and floats over the
cellar floor without contacting the ramp.
Next session pickup options:
1. Debug the airborne-at-tick-1 issue (likely in TransitionTypes
FindEnvCollisions indoor BSP path — why does a flat (0,0,1) hit
return walkable=False?). Once fixed, the harness should reproduce
cellar-up freeze.
2. Pivot to a different M1.5 issue with cleaner reproduction.
3. Use the harness mechanics elsewhere — the synthetic-GfxObj +
ShadowEntry pattern is reusable for any indoor-static-collision
test (corpse pickup boundaries, door swings, etc.).
Test baseline: 1167 + 5 (harness) = 1172 + 8 pre-existing failures.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Extends the trajectory replay harness with a programmatic mini-stair
piece, reconstructed from the live capture's polydump data
(docs/research/2026-05-21-a6-captures/scen4_cottage_cellar_polydump).
NEW finding: the cellar ramp polygon is NOT in cellStruct.PhysicsPolygons.
It lives in a separate GfxObjPhysics (the cellar's stair-piece static
building) registered via ShadowObjectRegistry, queried via
FindObjCollisions → engine.DataCache.GetGfxObj. CellDumpSerializer is
CORRECT — it captures the cell's physics polygons accurately. The
ramp polygon comes from a different data source entirely.
The polydump probe at BSPQuery.AdjustSphereToPlane:402 reports
"cell=0xA9B40147 polyId=0x0008 sides=Landblock" because the SPHERE
is in that cell at hit time — but the polygon's actual source is the
building's GfxObj. Inside the cellar fixture, polyId=0x0008 happens
to be a wall (Normal=(1,0,0)); inside the building's GfxObj, polyId
=0x0008 is the ramp (Normal=(0,-0.719,0.695) local). Same ID, different
collection.
The new RegisterStairRampGfxObj() in the harness constructs the
building's ramp polygon in WORLD coordinates (translated from
local building frame + 180° yaw), wraps it in a minimal one-leaf
PhysicsBSPTree, registers via cache.RegisterGfxObjForTest, and
attaches a ShadowEntry with cellScope=CellarId so the shadow is only
queried when the sphere is in the cellar cell (matches retail's
per-cell shadow scoping for interior statics — Issue #91 family).
Verified: world plane n=(0,0.719,0.695), d=-69.5035 (matches live
cdb capture exactly to 4 sig figs). Ramp foot at world Y=8.745,
Z=90.955; ramp top at world Y=5.845, Z=93.955. 3.0 m vertical rise.
NEW blocker discovered: the sphere goes airborne at tick 1 (same
issue documented in the prior commit's Finding #2). Sphere FLOATS
at Z=91.43 over the cellar floor, never contacts the synthetic
ramp. The synthetic stair registration mechanics are validated (the
GfxObj is in the cache, the ShadowEntry is in the registry, the
BSP tree is well-formed) — but trajectory replay still blocked on
the seeded-grounded-state bug. Next session needs to diagnose
WHY the engine reports "hit=yes n=(0,0,1) walkable=False" on tick 1
for a sphere correctly seeded as grounded on the cellar floor.
Test baseline maintained: 1167 + 4 (harness) = 1171 + 8 pre-existing
failures.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds tests/AcDream.Core.Tests/Physics/CellarUpTrajectoryReplayTests.cs:
a deterministic harness that drives PhysicsEngine.ResolveWithTransition
through N ticks against pre-loaded cell fixtures, capturing per-tick
trajectory points. Pure indoor (no landblock registration needed),
runs 200 ticks in under 100 ms.
The harness MECHANICS work — engine constructs cleanly, DataCache
accepts test fixtures via RegisterCellStructForTest, PhysicsBody
carries ContactPlane state across ticks. 4/4 tests pass, baseline
maintained (1167 + 4 = 1171 + 8 pre-existing failures).
Two real findings surfaced during commissioning, both documented as
passing tests so they don't regress silently:
Finding 1 (Harness_FixtureLimitation_NoRampPolygon): the three
issue-#98 cell fixtures contain ONLY axis-aligned polygons. The
cellar fixture (0xA9B40147) has 37 polys: 8 floor (N=(0,0,1)), 7
ceiling (N=(0,0,-1)), 22 walls. The live capture's CELLAR RAMP
polygon (N ≈ (0, ±0.719, 0.695)) is NOT in any fixture. With no
ramp polygon, the harness can't reproduce the cellar-up climb —
the sphere would walk horizontally across the cellar floor without
ever encountering a slope. Re-capture needed; investigate whether
CellDumpSerializer is skipping polygons or whether the ramp lives
in a cell we didn't dump.
Finding 2 (Harness_Finding_SphereGoesAirborneAtTick1): at the
seeded grounded initial position (sphere center 0.48 m above cellar
floor, ContactPlane = (0,0,1,-90.95), OnWalkable bit set), the
engine reports `hit=yes n=(0,0,1) walkable=False` on tick 1 and
the body's IsOnGround flips to false. Subsequent ticks proceed as
airborne (Y advances, Z stays put — no gravity in the input offset).
Unclear whether this is an engine bug (floor contact classified
as non-walkable collision) or a fixture issue (cellar floor
polygon's containment test mis-firing at the seeded XY). Either
way, the harness now exposes it deterministically.
Net value of this commit: the harness CODE is ready. Once the
fixture issue is solved, fix attempts on #98 (or any trajectory-
dependent bug) iterate in <100 ms instead of 5-minute live-launch
cycles. The "why is this so hard" point #4 from the session-pause
handoff is addressed for everything except the missing-ramp gap.
Test baseline: 1171 (1167 + 4 new) + 8 pre-existing failures.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Updates ISSUES.md and CLAUDE.md to reflect the actual state of #98
after two days of work:
- The new [step-walk-adjust] probe (8a232a3) + capture + findings
(8daf7e7) prove AdjustOffset's slope projection is CORRECT. Sphere
Z climbs monotonically 90.95 -> 92.80 across the ramp at +0.045 m
mean zGain per call. The earlier "Fix targets 1-4" priority list
is OBSOLETE — AdjustOffset is not the bug.
- The climb caps at world Z ~= 92.80 because step-up's downward
step-down probe finds no walkable within stepDownHeight=0.6 m
below the proposed position. Cottage floor at Z=94 is ABOVE, not
below. 101 stepdown-reject hits in the capture vs 1 acceptance.
- Shape 1 fix attempted (0cb4c59): gated BSPQuery.AdjustSphereToPlane's
two SetContactPlane call sites by Normal.Z >= 0.99 to match retail's
cdb-observed flat-CP-only pattern. Reverted (402ec10) — gate broke
OnWalkable tracking. 74% of new capture in falling state. User
report: "can't get up the first step, jumped, stuck in falling
animation." Either retail synthesizes a flat CP from sloped
contacts (step_sphere_down:321203 path, unclear from BN decomp)
or our OnWalkable tracking is over-coupled to ContactPlaneValid.
Apparatus state: probe, findings, replay harness, plan, retail
cdb capture all committed and ready for next session.
Honest next-session moves (in order):
1. Build deterministic trajectory replay harness — 200ms inner loop
instead of 5-minute live test. Issue98 cell fixtures are half of
this already.
2. Pivot to less-coupled M1.5 issue while #98 awaits the harness.
3. Deeper named-decomp research on CEnvCell::find_env_collisions ->
BSPTREE::find_collisions indoor CP-setting chain. Prior passes
worked on the outdoor CLandCell path; indoor was never traced.
NO further #98 fix attempts until apparatus or research has
converged. Five+ failed attempts in the saga is the signal.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds PhysicsGlobals.ContactPlaneFlatThreshold = 0.99f and uses it at
both BSPQuery.AdjustSphereToPlane call sites that previously set CP
unconditionally on any walkable polygon found by FindWalkableInternal.
Backed by the retail cdb capture in cellar_up_capture_1: across 161
set_contact_plane writes during 5 seconds of cellar-up climbing,
EVERY write lands on a flat (Normal.Z = 1.0) plane — cellar floor at
world Z=90.95 or cottage floor at world Z=94. The cellar ramp
(Normal.Z = 0.695, walkable per FloorZ but sloped ~46 degrees) is
never set as CP in retail.
Acdream's prior behavior of setting CP=ramp caused two cascading
issues at the top of the ramp:
1. AdjustOffset's slope-projection produced +Z gain per call (correct
in isolation) but inflated step-up's responsibility to "find the
next walkable below the lifted check position".
2. step-up's downward step-down probe found no walkable within 0.6m
below the proposed check (cottage floor at Z=94 is ABOVE, not
below), so step-down rejected, sphere rolled back. Infinite freeze
at world Z ~= 92.80.
With CP only set on flat polygons, sloped surfaces drive collision
detection and walkable-poly tracking (via path.SetWalkable) but
don't override the resting CP. The sphere should now climb the ramp
via step-up over the ramp polygon, with CP staying on the flat
cellar floor until the sphere reaches the flat cottage floor.
Tests: 1167 + 8 baseline maintained. No regression. The Issue98
replay tests still pass — they document the failing-frame geometry
(sphere world Z=92.01 below cottage floor), which doesn't change;
the fix prevents the sphere from getting STUCK at that altitude in
the first place. Live visual verification required next.
If the live test shows new failure modes (sphere stuck somewhere
else, doesn't climb at all, climbs but slides off, etc), the
threshold (0.99) or the gating approach itself may need refining.
This is the conservative empirical version of Shape 1; the named-
decomp research did not conclusively prove the exact retail gate.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The diagnostic-first capture revealed the failure mode the plan's
four-branch decision tree (A/B/C/D) did not anticipate. AdjustOffset
is CORRECT: 145/146 calls use the into-plane branch, mean zGain
+0.045 m per call, sphere world Z climbs 90.95 -> 92.80 monotonically.
The climb caps at world Z 92.80 (cottage floor at 94.00 is still
1.20 m above). At the cap, the per-step CP reset at TransitionTypes.cs
723-725 clears ContactPlaneValid as designed; TransitionalInsert
should re-establish CP at the proposed position. Step-up logic fires
because the offset has +Z; step-up calls DoStepDown(stepDownHeight=
0.6, runPlacement=true). The downward probe finds NO walkable surface
within 0.6 m below the proposed position (cottage floor is ABOVE,
not below) -- 101 stepdown-reject hits in this capture vs 1 acceptance.
Conclusion: Target E (new). Three candidate fix shapes named in the
findings note. Each one researched against retail named-decomp before
any code lands. Test baseline 1167 + 8 maintained.
Findings: docs/research/2026-05-23-a6-stepwalkadjust-findings.md
Capture: docs/research/2026-05-23-a6-captures/stepwalkadjust/acdream.log
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds one log line per AdjustOffset call (gated by ACDREAM_PROBE_STEP_WALK)
naming the branch taken (no-cp / no-cp-slide / slide-degenerate /
slide-crease / into-plane / away-plane, optionally +safety-push) plus
zGain = output.Z - input.Z.
No math or control-flow changes — pure observability so the next capture
can disambiguate the three failure-mode hypotheses for the cellar-ramp
climb cap. Re-reading the existing capture (a6-issue98-negpoly-...log)
showed the sphere DOES climb 90.00 -> 92.79 (2.79 m gain), then caps,
contradicting the divergence comparison's "no altitude gain" framing.
The real question is what stops the climb at world Z ~= 92.79 with the
cottage floor still 1.21 m higher. Existing [step-walk] probes wrap
AdjustOffset; this new probe reveals which branch the projection takes.
Fix plan with the four-branch decision tree at
docs/superpowers/plans/2026-05-23-a6-p3-issue98-cellar-up-fix.md.
Test baseline maintained: 1167 + 8.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Final step of the apparatus plan. Updates ISSUES.md issue #98 and
CLAUDE.md's M1.5 status to reflect:
- The apparatus completed (Steps 1-5 land in commits 35b37df →
28c282a).
- The real divergence: retail's sphere is at world Z ≈ 94.48 (resting
on cottage floor) when find_walkable accepts; acdream's failing-
frame sphere is 2.47m lower at world Z ≈ 92.01.
- The four fix targets, in priority order. Fix plan is the NEXT plan,
scoped to Target 1 (step-up + ramp climb Z gain) or Target 2
(cottage-cell sphere reference).
- The replay harness (Issue98CellarUpReplayTests) is the test loop —
any fix that doesn't change the failing assertions is not the fix.
Today's commit graph on top of slice 5 (cf3deff):
35b37df triage — revert neg-poly + bldg-check experiments
f62a873 Step 2 — cell-dump probe + roundtrip test
3f56915 Step 2 capture — 3 real-geometry cell fixtures
856aa78 Step 3 — deterministic replay harness (7 tests)
6f666c1 Step 4 — retail cdb find_walkable capture script
28c282a Step 5 — replay vs retail divergence comparison
(this) Step 6 — ISSUES.md + CLAUDE.md handoff
Test baseline: 1167 + 8 (8 pre-existing failures, +19 new passing
tests across the apparatus). Build green throughout.
A6.P3 #98 is now in evidence-driven mode. Fix plan starts from the
divergence doc at
docs/research/2026-05-23-a6-p3-issue98-replay-comparison.md.
Pickup prompt for the fix-plan session is in §"Pickup prompt for the
fix plan" of that doc.
Closes the apparatus loop. Side-by-sides acdream's deterministic replay
(commit 856aa78) against retail's cdb capture taken via Step 4's
runner. The divergence target is named; the fix plan is the next plan.
Retail data (cellar_up_capture_1):
- 35,219 BP hits over ~5 seconds of motion
- BPE (set_contact_plane): 161 writes, ALL to one of two flat planes
(n=(0,0,1) d=-93.9998 = cottage floor @ Z=94, OR d=-90.95 = cellar
floor @ Z=90.95). Retail NEVER sets ContactPlane to the cellar ramp.
- BPC (find_crossed_edge): 1 hit in 35K. Retail barely uses this
predicate during cellar-up.
- BPA (find_walkable) sphere position at each cottage-floor
acceptance: sphere LOCAL Z = +0.48 to +0.63 (resting on top of the
floor plane). Sphere world Z ≈ 94.48.
acdream replay (Issue98CellarUpReplayTests):
- At the failing-frame sphere (world (141.7, 8.4, 92.0)), the cottage
cell 0xA9B40143's poly 0x0004 reports insideEdges=false AND
overlapsSphere=false. Sphere local Z = -0.69 (below the cottage
floor plane). 0xA9B40146 has no walkable candidate at all. Step-up
has nothing to step onto → stuck.
Sphere world Z delta: 2.47m. Retail's sphere is 2.5m higher than ours
at the decision point. The fix targets, in priority order:
1. (HIGHEST CONFIDENCE) Step-up + ramp climb doesn't gain enough Z per
tick. Retail climbs the ramp GRADUALLY across thousands of ticks;
ours oscillates at world Z ≈ 92 without altitude gain. Look at
Transition.AdjustOffset (slope projection) and Transition.DoStepUp
(does it reset WalkInterp like retail's step_sphere_up?).
2. Cottage-cell candidacy uses wrong sphere reference. Check what
sphere CheckOtherCells passes to BSPQuery.FindCollisions — is it
the step-lifted sphere or the pre-step sphere?
3. (SECONDARY) find_crossed_edge over-use. Our walkable test calls
FindCrossedEdge heavily; retail barely uses it. Possibly a
code-shape mismatch in step-up vs walkable-acceptance flow.
4. (LOW CONFIDENCE) Ramp polygon normal divergence. Verify via test
after any fix.
The apparatus that gets us here:
- tests/AcDream.Core.Tests/Fixtures/issue98/*.json (real cell geometry)
- Issue98CellarUpReplayTests (7 tests, <1ms each, deterministic bug
reproduction)
- tools/cdb/issue98-runner.ps1 (reusable for any future capture)
- docs/research/2026-05-23-a6-captures/cellar_up_capture_1/ (this
capture, checked in for future analyses)
Next plan: pick Target 1 or 2 from the comparison doc and write the
fix plan against it. The replay harness is the test loop; a fix that
doesn't change the failing assertions in Issue98CellarUpReplayTests is
not the fix.
Step 4 of the apparatus plan. Adds the cdb script + runner that pairs
with Issue98CellarUpReplayTests to compare retail's walkable-query
behavior against acdream's during the Holtburg cottage cellar ascent.
Breakpoints (all symbols verified against refs/acclient.pdb via grep
docs/research/named-retail/symbols.json):
- BPA: BSPLEAF::find_walkable — leaf-level walkable query
- BPB: CPolygon::walkable_hits_sphere — per-polygon overlap test
- BPC: CPolygon::find_crossed_edge — per-polygon edge containment
- BPD: CTransition::check_other_cells — outer dispatcher
- BPE: COLLISIONINFO::set_contact_plane — GOLD signal: retail accepted
this plane
- BPF: CPolygon::adjust_sphere_to_plane — per-polygon projection
Output format: 32-bit hex bits for all floats via dwo() + %08X (cdb's
%f handling is broken for dwo reads; see a6-probe.cdb v3→v4 history).
Decoder: tools/cdb/decode_retail_hex.py already handles _h=0x... fields.
Auto-detach threshold: 50000 hits across BPA/B/C/D/F. BPE is unbounded
(contact plane writes are rare, ~18 per ascent per slice 5 capture).
Runner: tools/cdb/issue98-runner.ps1
.\tools\cdb\issue98-runner.ps1 -ScenarioTag "cellar_up_attempt_1"
Prereqs (per CLAUDE.md retail debugger toolchain section):
- Retail acclient.exe v11.4186 running and in-world
- ACE running on 127.0.0.1:9000
- Character at the BOTTOM of a Holtburg cottage cellar stair
- cdb.exe present at the Windows Kits 10 path
Output:
docs\research\2026-05-23-a6-captures\<ScenarioTag>\retail.log
Reading the log:
- [BPE] lines tell you which plane retail accepted (the answer we need).
- Cross-reference [BPE]'s normal/d against the cell fixtures in
tests/AcDream.Core.Tests/Fixtures/issue98/*.json to identify which
cell + polyId retail picked.
- The divergence between retail's accepted polygon and our replay test's
"no walkable accepted" result IS the fix target.
The capture itself is a user action (cdb requires a live retail
process); this commit only ships the protocol. Step 5 (comparison doc)
follows after the capture lands.
Step 3 of the apparatus plan. Adds Issue98CellarUpReplayTests, a 7-test
harness that loads the three real-geometry cell fixtures captured in
commit 3f56915 and drives the failing-frame sphere through the same
nearest-walkable algorithm the live client uses in
Transition.LogNearestWalkableCandidate.
The tests reproduce the live failure deterministically in under 1ms
each — the issue #98 cellar-up bug is now visible to a unit-test run,
no client launch required.
Tests:
- Fixtures_AllThreeCellsLoadAndShareOrigin — sanity check the cells
loaded with the expected (130.5, 11.5, 94.0) origin.
- Cellar_HasMostPolygons_CottageNeighborBIsSparse — confirms the
surprising finding: 0xA9B40146 is too sparse to be a "cottage main
floor" cell (slice 5 handoff inference was wrong; 0xA9B40143 with 14
polys is the better candidate).
- FailingFrame_CellarPrimary_HasCellarRampAsNearestWalkable — the
ramp polygon IS reachable when the player is on top of it
(sanity: this should always be true).
- FailingFrame_CottageNeighborA_NearestWalkableIsOutsideSphereAndEdges
— at the failing-frame sphere position, the nearest walkable in
0xA9B40143 (poly 0x0004, the cottage floor triangle at world Z=94)
reports BOTH insideEdges=false AND overlapsSphere=false. The sphere
XY is beyond the triangle edge, and the sphere is too far below the
plane. THIS IS THE BUG'S SHAPE.
- FailingFrame_CottageNeighborB_HasNoWalkableCandidate — 0xA9B40146
has NO walkable polygon close enough to the failing-frame sphere.
- FailingFrame_NoCottageNeighbourYieldsAcceptedWalkable — composite:
across both cottage cells, no walkable passes both edge + sphere
tests → step-up has nothing to step onto → player stuck.
- FailingFrame_CottageNeighborA_Poly0x0004_HasExpectedShape — pins the
exact polygon shape so a future fixture re-capture failure is loud.
What this gives us:
1. The bug is now ALWAYS reproducible in test, no live client iteration.
2. Any fix to BSPQuery.FindCrossedEdge / polygon containment / the
cell transform will instantly show whether it changes the failing-
frame outcome.
3. Step 4 (retail cdb capture) will tell us what retail finds at the
same sphere position; Step 5 (comparison doc) will name the
divergence; the eventual fix is then evidence-driven, not a guess.
The tests document the CURRENT (failing) behavior. They WILL pass
after the fix — at which point they need to flip to assert the
retail-correct behavior. This intentional brittleness is the point:
the test is the bug's gravestone, and a fix that doesn't match retail
should not satisfy the test.
Verification:
- dotnet build: green, 0 errors.
- dotnet test: 1167 passed + 8 pre-existing failed (was 1160+8 before
this commit; +7 from the replay tests). Same pre-existing failures,
no new regressions.
- Each Issue98 test runs in under 1ms; loads JSON, calls one internal
predicate per polygon, asserts.
Next: tools/cdb/issue98-cellar-up-find-walkable.cdb (Step 4).
Step 2 capture step. Launched the live client with
ACDREAM_DUMP_CELLS=0xA9B40143,0xA9B40146,0xA9B40147 and walked into a
Holtburg cottage cellar. The probe fired on first-cache of each cell
and emitted JSON dumps to tests/AcDream.Core.Tests/Fixtures/issue98/.
Cell contents (resolved polygons + portals):
- 0xA9B40143: 14 polys + 4 portals (~18.7 KB)
- 0xA9B40146: 4 polys + 2 portals (~7.0 KB)
- 0xA9B40147: 37 polys + 2 portals (~45.7 KB) — cellar, biggest
All three share worldOrigin=(130.5, 11.5, 94.0) with 180° yaw rotation
(M11=M22=-1), matching the failing-frame's local-to-world projection.
Reproduction during capture: spawn at (141.6, 8.4, 91.5) @ 0xA9B40147
— almost exactly the slice 7 handoff's failing-frame position. User
tried to walk up the cellar stair and got stuck (issue #98 reproduction
confirmed).
Surprise: 0xA9B40146 with only 4 polys + 2 portals is too sparse to be
the "cottage main floor cell" that the slice 5 handoff inferred — that
designation was a guess, not verified. 0xA9B40143 (14 polys) is the
better candidate. Step 3 (replay harness) will confirm by inspecting
the actual polygon geometry against the failing-frame sphere position.
Cells are real geometry from client_cell_1.dat, not synthetic fixtures.
The replay harness can now drive the leaf-level walkable predicates on
this exact data without launching a window.
Next: Issue98CellarUpReplayTests (Step 3).
