fix(physics): AP-152 — dispatch collision shapes BSP-first, at emission and at the cell flood

The register row predicted "catching or stopping on a doorway sill". That
symptom could not have been occurring. `Transition.BspOnlyDispatch`
(TransitionTypes.cs:1348, landed 2026-05-25 as A6.P7) already skipped both
primitive branches (:3911, :3954) whenever the target's wire PhysicsState
carries HAS_PHYSICS_BSP_PS, and ACE sets that bit from CSetup.HasPhysicsBSP
for every affected Setup. The extra primitive was never tested for collision.

The live defect was CELL MEMBERSHIP. The same shape list feeds
`ShadowObjectRegistry.BuildFloodSpheres`, which had no such guard and
preferred Cylinders over everything whenever any Cylinder existed — retail's
SECOND priority applied ahead of its first. For the 73 CylSphere+BSP Setups
acdream therefore flooded shadow cells from the cylinder and never from the
slab: an object absent from cells it physically occupies, which is the
#98 / #168 symptom class, not the door-collision class the row named.

Retail, re-disassembled from the PDB-paired binary (v11.4186, CodeView GUID
9e847e2f-777c-4bd9-886c-22256bb87f32, check_exe_pdb.py MATCH) rather than
taken from Binary Ninja, which drops flag tests:

  CPhysicsObj::FindObjCollisions @0x0050f050
    0x0050f165  test dword [esi+0xa8], 0x10000
    0x0050f16f  je   0x50f1a2        ; clear -> primitive dispatch
    0x0050f18d  call 0x518180        ; CPartArray::FindObjCollisions
    0x0050f19d  jmp  0x50f2b0        ; UNCONDITIONAL, past BOTH primitive loops
                                     ; (CylSphere 0x50f1a2, Sphere 0x50f21d)
    0x0050f1d6  jae  0x50f317        ; CylSphere loop exhausted -> RETURN
    0x0050f22f  je   0x50f31b        ; zero Spheres -> RETURN seeded OK_TS

  CPhysicsObj::calc_cross_cells @0x00515230
    0x00515285  test dword [esi+0xa8], 0x10000
    0x0051528f  jne  0x515305 -> CPhysicsObj::find_bbox_cell_list @0x00510fc0
    0x005152d1  call 0x52b9f0        ; cylsphere branch, below the jump
    0x005152fb  call 0x52b990        ; sorting-sphere branch, below the jump

Priority at both consumers: BSP -> CylSphere -> Sphere -> nothing. BSP wins.
Every address above was resolved back to its symbol by exact lookup in
named-retail/symbols.json.

Changes:

* `ShadowShapeBuilder.FromSetup` gains a step-0 dispatch gate. Steps 1 and 2
  are skipped entirely when any part's EFFECTIVE GfxObj carries a physics
  BSP. The gate and step 3 now share one `EffectivePartGfxObjId` helper, so
  they cannot read different identities — a gate on `setup.Parts` would,
  after an ObjDesc swap, suppress the primitives while step 3 emitted
  nothing and `Build` returned null, deleting the entity's collision.
  Emission order is unchanged. This also removes acdream's undeclared
  reliance on the server sending the flag: the gate is derived from the
  parts, exactly as CPartArray::CacheHasPhysicsBSP @0x00518110 derives it.

* `ShadowObjectRegistry.BuildFloodSpheres` now applies calc_cross_cells'
  own order: BSP, else Cylinder, else everything. Given the gate above this
  is a no-op for every shape list acdream produces (FromSetup is now
  exclusive; both landblock-static publishers already emit homogeneous
  lists), so the measured membership delta remains attributable to the
  gate alone. It is kept for the same reason BspOnlyDispatch is kept: retail
  genuinely dispatches here, and it guards a future additive producer.

`Transition.BspOnlyDispatch` is deliberately untouched.

Register: AP-152 RETIRED with its four false statements corrected — the risk
statement (the symptom was already inert); "small and centred at the part
origin" (max primitive is 6.714 m, and 0x0200086E's sphere origin is
(0.759, 0.165, 5.842)); the cottage door's "~14 cm base Sphere" (it is
0.100 m; 0.141 is Setup.Radius, which AP-22 proved is never collision
geometry); and naming one pinning test where two existed. AP-153/154/155
filed: retail's dispatch flag is cached once at InitPartArrayObject+0x7e
where acdream's gate is live; the query-time guard takes a client-derived
flag off the wire; and the static publishers emit Setup Spheres as
height-capped Cylinders while BuildFloodSpheres approximates retail's
bounding box with bounding spheres.

Tests. Both pinning tests corrected, neither deleted:
`FromSetup_DoorSetup_ProducesFourShapes` -> `..._EmitsBspPartsOnly`;
`FromSetup_DoorSetup_SphereAtExpectedLocalOffset` re-hosted on
`_ => false`, the DAT-real configuration for the 3,605 Sphere-only Setups.
`FromSetup_ScaleFactor_MultipliesAllRadiiAndOffsets` was the campaign's
eighth green test covering nothing — its assertions sat inside
`if (CollisionType == Cylinder)` on a fixture with zero CylSpheres, so only
`Scale == 2.0f` ever ran. Proved empirically: with the sphere radius scale
deleted, the old body passes and the corrected body fails. Three new facts:
the effective-identity gate, the App-layer CylSphere+BSP registration (no
App fixture combined the two before), and the flood-set dispatch. One new
installed-DAT sweep pins 172 affected Setups (73 CylSphere+BSP, 99
Sphere+BSP) behind external bucket controls, re-measured independently and
agreeing exactly with the filing commit's separate sweep.

All eight sabotages run and reported; every discriminating fact reddens in
the intended direction and only there. Clean Release build after deleting
every bin/obj: 0 errors. Complete suite 11,203 passed / 4 skipped / 0
failed, +5 on the 11,198 baseline at ec29a732 — exactly the five added
facts, no new skips.

Blast radius, corrected: the FromSetup half is graphical-only (its sole
production caller is LiveEntityCollisionBuilder in AcDream.App, which
AcDream.Headless cannot reference — Headless -> Runtime -> Core/Content).
The BuildFloodSpheres half lives in AcDream.Core and DOES execute in
Headless via LandblockPhysicsContentBuilder, but is behaviour-neutral there
because both of that builder's registrations pass homogeneous lists.
Headless suite green at 89/89.

NOT yet gated live: this changes shadow-cell membership for 22 Setups used
by 151 Door weenies and 38 stationary props. Needs a connected session.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
Erik 2026-08-06 14:51:36 +02:00
parent ec29a732f5
commit 4abd1b5eb7
9 changed files with 1635 additions and 85 deletions

View file

@ -597,11 +597,35 @@ public sealed class ShadowObjectRegistry
}
/// <summary>
/// Retail flood-sphere rule (CylSphere overload, Ghidra 0x0052b9f0):
/// when the object has cylinder shapes, each contributes one sphere at
/// its world BASE point (low_pt) with the cylinder radius, capped at 10;
/// otherwise the BSP parts' bounding spheres are the footprint (the
/// sorting-sphere fallback, calc_cross_cells 0x00515230 tail).
/// Retail cross-cell dispatch, <c>CPhysicsObj::calc_cross_cells</c>
/// @0x00515230, in retail's own priority order:
///
/// <list type="number">
/// <item>BSP-bearing (<c>0x00515285 test dword [esi+0xa8],0x10000</c> /
/// <c>0x0051528f jne 0x515305</c>) → <c>CPhysicsObj::find_bbox_cell_list</c>
/// @0x00510fc0. The cylsphere and sorting-sphere branches are BOTH below
/// that jump and unreachable from it. acdream approximates the bbox with
/// the BSP parts' bounding spheres.</item>
/// <item>else cylspheres (<c>0x00515298 GetNumCylsphere</c> non-zero) →
/// <c>CObjCell::find_cell_list</c> @0x0052b9f0 over the cylsphere array;
/// each contributes one sphere at its world BASE point with the cylinder
/// radius, capped at 10.</item>
/// <item>else the sorting sphere (<c>0x005152dc</c> →
/// <c>CPartArray::GetSortingSphere</c> @0x00518b00 →
/// <c>CObjCell::find_cell_list</c> @0x0052b990).</item>
/// </list>
///
/// <para>
/// The BSP-first rule is redundant for every shape list acdream produces
/// today — <see cref="ShadowShapeBuilder.FromSetup"/> dispatches at
/// emission (AP-152) and both landblock-static publishers emit
/// homogeneous lists — exactly as
/// <c>Transition.BspOnlyDispatch</c> is redundant at the query site. It is
/// kept because retail genuinely dispatches here, and because a producer
/// that handed this method a mixed list would otherwise flood a
/// BSP-bearing object from its primitive and silently place it in the
/// wrong shadow cells (the #98 / #168 symptom class).
/// </para>
/// </summary>
private static List<DatReaderWriter.Types.Sphere> BuildFloodSpheres(
Vector3 entityWorldPos,
@ -611,15 +635,25 @@ public sealed class ShadowObjectRegistry
const int RetailSphereCap = 10;
var spheres = new List<DatReaderWriter.Types.Sphere>();
bool anyBsp = false;
bool anyCyl = false;
foreach (var s in shapes)
{
if (s.CollisionType == ShadowCollisionType.Cylinder) { anyCyl = true; break; }
if (s.CollisionType == ShadowCollisionType.BSP) anyBsp = true;
else if (s.CollisionType == ShadowCollisionType.Cylinder) anyCyl = true;
}
// Retail's branch, chosen once: BSP-bbox, else cylspheres, else the
// sorting sphere (which acdream approximates with the remaining
// shapes' bounding spheres).
ShadowCollisionType? only =
anyBsp ? ShadowCollisionType.BSP
: anyCyl ? ShadowCollisionType.Cylinder
: null;
foreach (var s in shapes)
{
if (anyCyl && s.CollisionType != ShadowCollisionType.Cylinder)
if (only is { } required && s.CollisionType != required)
continue;
if (spheres.Count >= RetailSphereCap)
break;

View file

@ -14,21 +14,25 @@ namespace AcDream.Core.Physics;
/// <see cref="ShadowObjectRegistry.RegisterMultiPart"/>.
///
/// <para>
/// Walks (1) every CylSphere → Cylinder shape, (2) every Sphere ONLY when no
/// CylSpheres are present (matches retail and the existing landblock-static
/// convention at GameWindow.cs:6034), and (3) every Part whose GfxObj has a
/// non-null PhysicsBSP → per-part BSP shape, with local transforms from
/// PlacementFrames[Resting | Default | first available].
/// This is a DISPATCH, not a union. In priority order: (3) when ANY Part's
/// effective GfxObj has a non-null PhysicsBSP, emit one BSP shape per such
/// Part — and nothing else; otherwise (1) every CylSphere → Cylinder shape;
/// otherwise (2) every Sphere → Sphere shape; otherwise nothing at all. Local
/// transforms come from PlacementFrames[Resting | Default | first available].
/// </para>
///
/// <para>
/// Retail anchor: <c>CPhysicsObj::FindObjCollisions</c> (0x0050f050)
/// dispatches EXCLUSIVELY on <c>HAS_PHYSICS_BSP_PS</c> (0x10000): it calls
/// dispatches EXCLUSIVELY on <c>HAS_PHYSICS_BSP_PS</c> (0x10000)
/// (<c>0x0050f165 test dword [esi+0xa8],0x10000</c> /
/// <c>0x0050f16f je 0x50f1a2</c>): it calls
/// <c>CPartArray::FindObjCollisions</c> (the per-part BSP walk) and returns
/// (<c>0x0050f19d jmp</c> past the primitive branches), OR walks the Setup's
/// CylSpheres, OR walks the Setup's Spheres, OR — with none of the three —
/// returns the seeded <c>OK_TS</c> without synthesizing any shape
/// (<c>0x0050f22f je 0x50f31b</c>). It is never a union.
/// (<c>0x0050f19d jmp 0x50f2b0</c>, an UNCONDITIONAL jump past both primitive
/// branches — the CylSphere loop starts at 0x50f1a2 and the Sphere loop at
/// 0x50f21d), OR walks the Setup's CylSpheres and returns
/// (<c>0x0050f1d6 jae 0x50f317</c>), OR walks the Setup's Spheres, OR — with
/// none of the three — returns the seeded <c>OK_TS</c> without synthesizing
/// any shape (<c>0x0050f22f je 0x50f31b</c>). BSP wins.
/// <c>CPhysicsPart::find_obj_collisions</c> (0x0050d8d0) tests ONLY the
/// GfxObj physics BSP; CylSpheres and Spheres are <c>Setup</c>-level arrays
/// reached through <c>CPartArray::GetCylsphere</c> (0x00518090) and
@ -36,12 +40,28 @@ namespace AcDream.Core.Physics;
/// </para>
///
/// <para>
/// KNOWN DIVERGENCE (AP-152): steps 1/2 and step 3 below are emitted
/// ADDITIVELY here, where retail is exclusive — 172 of 5,935 installed
/// Setups carry a primitive and a physics-BSP part. Do not cite the anchor
/// above as justification for the additive design; it is the evidence
/// against it. acdream's two static publication paths already implement the
/// exclusive rule.
/// Cell membership dispatches on the SAME flag and in the same priority:
/// <c>CPhysicsObj::calc_cross_cells</c> (0x00515230) tests
/// <c>0x10000</c> at <c>0x00515285</c> and routes a BSP-bearing object to
/// <c>CPhysicsObj::find_bbox_cell_list</c> (0x00510fc0) at
/// <c>0x0051528f jne 0x515305</c>, never reaching its cylsphere
/// (<c>0x005152d1</c>) or sorting-sphere (<c>0x005152fb</c>) branches. That
/// is why the exclusivity is enforced HERE, at emission, rather than only at
/// the query-time guard <c>Transition.BspOnlyDispatch</c>: the shape list is
/// also the input to <c>ShadowObjectRegistry.BuildFloodSpheres</c>.
/// </para>
///
/// <para>
/// AP-152 (filed and retired 2026-08-06): these three steps used to be
/// emitted ADDITIVELY — 172 of 5,935 installed Setups carry both a primitive
/// and a physics-BSP part. The collision half of that divergence was already
/// inert, because <c>Transition.BspOnlyDispatch</c> skips both primitive
/// branches whenever the wire <c>PhysicsState</c> carries 0x10000 and ACE
/// derives that bit from the same DAT flag; the live half was CELL
/// MEMBERSHIP, which had no such guard. Gating here also removes acdream's
/// undeclared dependency on the server sending the bit: the gate is derived
/// from the parts, exactly as retail's <c>CPartArray::CacheHasPhysicsBSP</c>
/// (0x00518110) derives it.
/// </para>
/// </summary>
public static class ShadowShapeBuilder
@ -81,38 +101,66 @@ public static class ShadowShapeBuilder
var result = new List<ShadowShape>();
// 1. CylSpheres — each becomes a Cylinder shape.
foreach (var cyl in setup.CylSpheres)
// 0. Retail dispatch gate. CPhysicsObj::FindObjCollisions tests
// HAS_PHYSICS_BSP_PS FIRST (0x0050f165) and leaves the BSP branch
// through an unconditional jmp past both primitive loops
// (0x0050f19d); CPhysicsObj::calc_cross_cells tests the same flag
// at 0x00515285 and routes to find_bbox_cell_list. Retail derives
// the flag from the part array itself
// (CPartArray::CacheHasPhysicsBSP 0x00518110 ORs 0x10000 on the
// first part whose gfxobj->physics_bsp is non-null), so the gate
// below reads the SAME effective part identities step 3 reads —
// never setup.Parts directly. A gate keyed on a different identity
// could suppress the primitives while step 3 emitted nothing,
// silently deleting the entity's collision.
bool anyPhysicsBspPart = false;
for (int i = 0; i < setup.Parts.Count; i++)
{
if (cyl.Radius <= 0f) continue;
float baseHeight = cyl.Height > 0f ? cyl.Height : cyl.Radius * 4f;
result.Add(new ShadowShape(
GfxObjId: 0u,
LocalPosition: new Vector3(cyl.Origin.X, cyl.Origin.Y, cyl.Origin.Z) * entScale,
LocalRotation: Quaternion.Identity,
Scale: entScale,
CollisionType: ShadowCollisionType.Cylinder,
Radius: cyl.Radius * entScale,
CylHeight: baseHeight * entScale));
if (hasPhysicsBsp(EffectivePartGfxObjId(setup, effectivePartGfxObjIds, i)))
{
anyPhysicsBspPart = true;
break;
}
}
// 2. Spheres — only when no CylSpheres (matches landblock-static convention
// at GameWindow.cs:6034). Each becomes a true Sphere (no height clamping).
// Retail anchor: CSphere::intersects_sphere @ 0x00537A80 uses 3-D distance
// for the overlap check, unlike CCylSphere which clips to [low_pt, high_pt].
if (setup.CylSpheres.Count == 0)
// Steps 1 and 2 run ONLY for an object with no physics-BSP part.
if (!anyPhysicsBspPart)
{
foreach (var sph in setup.Spheres)
// 1. CylSpheres — each becomes a Cylinder shape.
foreach (var cyl in setup.CylSpheres)
{
if (sph.Radius <= 0f) continue;
if (cyl.Radius <= 0f) continue;
float baseHeight = cyl.Height > 0f ? cyl.Height : cyl.Radius * 4f;
result.Add(new ShadowShape(
GfxObjId: 0u,
LocalPosition: new Vector3(sph.Origin.X, sph.Origin.Y, sph.Origin.Z) * entScale,
LocalPosition: new Vector3(cyl.Origin.X, cyl.Origin.Y, cyl.Origin.Z) * entScale,
LocalRotation: Quaternion.Identity,
Scale: entScale,
CollisionType: ShadowCollisionType.Sphere,
Radius: sph.Radius * entScale,
CylHeight: 0f));
CollisionType: ShadowCollisionType.Cylinder,
Radius: cyl.Radius * entScale,
CylHeight: baseHeight * entScale));
}
// 2. Spheres — only when no CylSpheres. Retail's CylSphere loop
// returns rather than falling into the Sphere loop
// (0x0050f1d6 jae 0x50f317). Each becomes a true Sphere (no
// height clamping): CSphere::intersects_sphere @ 0x00537A80
// uses 3-D distance for the overlap check, unlike CCylSphere
// which clips to [low_pt, high_pt].
if (setup.CylSpheres.Count == 0)
{
foreach (var sph in setup.Spheres)
{
if (sph.Radius <= 0f) continue;
result.Add(new ShadowShape(
GfxObjId: 0u,
LocalPosition: new Vector3(sph.Origin.X, sph.Origin.Y, sph.Origin.Z) * entScale,
LocalRotation: Quaternion.Identity,
Scale: entScale,
CollisionType: ShadowCollisionType.Sphere,
Radius: sph.Radius * entScale,
CylHeight: 0f));
}
}
}
@ -126,10 +174,7 @@ public static class ShadowShapeBuilder
// degrade array before CPartArray::FindObjCollisions reads it.
// Keep the stable Setup part index/pose, but source collision
// identity from that effective part when one was supplied.
uint gfxId = effectivePartGfxObjIds is not null
&& i < effectivePartGfxObjIds.Count
? effectivePartGfxObjIds[i]
: (uint)setup.Parts[i];
uint gfxId = EffectivePartGfxObjId(setup, effectivePartGfxObjIds, i);
if (!hasPhysicsBsp(gfxId)) continue;
Frame partFrame;
@ -241,6 +286,20 @@ public static class ShadowShapeBuilder
return shapes;
}
/// <summary>
/// The collision identity of part <paramref name="index"/>: the installed
/// <c>AnimPartChanged</c> replacement when one was supplied, else the
/// Setup's own part. Shared by the step-0 dispatch gate and the step-3
/// emission so the two can never read different identities.
/// </summary>
private static uint EffectivePartGfxObjId(
Setup setup,
IReadOnlyList<uint>? effectivePartGfxObjIds,
int index)
=> effectivePartGfxObjIds is not null && index < effectivePartGfxObjIds.Count
? effectivePartGfxObjIds[index]
: (uint)setup.Parts[index];
/// <summary>Resolve the placement frame in priority Resting → Default →
/// first available. Mirrors <c>SetupMesh.Flatten</c>'s convention.</summary>
private static AnimationFrame? ResolvePlacementFrame(Setup setup)