A remote observed in acdream landed on a sloped roof and froze; the server slid
on, the gap passed AP-87's 4 m threshold, and the body snapped — the visible
blip. Live probe capture, two adjacent ticks 63 ms apart:
t=88420671 rsInContact=True rsOnWalkable=False rsIsOnGround=True
bodyCpNz=0.6097 floorZ=0.6642 steep=True gravity=True
vel=(2.146,2.264,-3.549)
t=88420734 contact=True onWalkable=True <- forced against the sweep
gravity=False <- cleared
velBeforeZero=(2.146,2.264,0.000)
moved=0.0000 <- and every tick after
The roof is 52.4 degrees against a 48.4 degree limit, so acdream's classifier
was CORRECT and was then overruled. Four independent links each froze the body
on their own: a per-tick force of Contact|OnWalkable, a per-tick velocity zero,
a Gravity clear at landing, and a landing edge testing IsOnGround
(= inContact || ...) instead of OnWalkable. The tick called
HandleAllCollisions alone — the tail of SetPositionInternal without its prefix.
Retail simulates remotes locally and derives these bits rather than asserting
them: CPhysics::UseTime @0x00509950 iterates the whole object table;
update_object @0x00515D10 gates only on parent/cell/FROZEN with no
is_player fork; SetPositionInternal @0x00515330 sets CONTACT from
contact_plane_valid @0x00515430 and ON_WALKABLE from contact_plane.N.z vs
floor_z @0x00515465-@0x0051548E before handle_all_collisions @0x005154FE;
set_on_walkable @0x00511310 fires HitGround @0x00511364 / LeaveGround
@0x00511346 edge-triggered with no ownership gate; calc_acceleration
@0x00510950 zeroes only when CONTACT && ON_WALKABLE && !Sledding @0x0051096B;
calc_friction @0x0050EE70 returns at its first line when ON_WALKABLE is clear.
acdream had copied retail's airborne no-op WITHOUT retail's local simulation.
The fix is mostly deletion: stop forging the transients, stop discarding the
authoritative velocity, stop clearing Gravity, and route the remote tick
through the same SetPositionInternal commit TickHidden and the local player
already use, with the landing edge derived from the sweep's own OnWalkable.
AP-87's threshold and conditions and InterpolationManager's node_fail_counter
snap-to-tail are deliberately untouched — this removes the CAUSE of the
divergence rather than weakening the backstop.
Cross-checked against ACE: its only creature-side VectorUpdate emitters are the
jump broadcast and spell projectiles, so integrating the wire velocity cannot
double-move a walking remote; and PhysicsGlobals.DefaultState already carries
Gravity, so deleting the manufactured State |= Gravity is safe.
Register: AP-81 narrowed (its GRAVITY half retired outright), AP-87 annotated,
AP-139 filed (the interpolation-queue clear on the landing edge), AP-140 filed
(the two routing gates select snap-vs-interpolate on walkability where retail
uses CONTACT — adjust_offset @0x00555D30 gates on transient_state & 1
@0x00555D52). AP-140's follow-up is deliberately shaped as "point the two gates
at Body.InContact", NOT "re-derive Airborne", which would perturb five writers
and collide with a pinned RemoteTeleportPlacementTests assertion.
Three gaps recorded in #32 rather than papered over: the new LeaveGround
dispatch is untested for chatter; a persistently !Ok transition can latch a
remote airborne; and — the visual-gate watch item — the deleted forge was a
blanket guarantee of Contact|OnWalkable, and contact_allows_move @0x00528dd0
silently refuses action animations without both, which is the literal root
cause of closed#270. Retail-correct on a steep face, a regression anywhere
else.
10 discriminating tests over a real PhysicsEngine landblock whose contact
normal Z is 0.61 against FloorZ 0.6642 — the live roof's exact relationship.
Suite 11,019 passed / 4 skipped / 0 failed. Includes the temporary
ACDREAM_PROBE_REMOTE_LANDING / ACDREAM_PROBE_REMOTE_SLIDE probe family that
produced the capture above; strip with the family.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Restore the named-retail object update order across local, remote, static, projectile, animation, shadow, teleport, and effect lifetimes. Separate authoritative root commits from spatial rebucketing, preserve per-owner hook/FIFO ordering, and remove update-path allocations with exact lifecycle and residency gates.
Add deterministic conformance, adversarial lifetime, GUID-reuse, pending-cell, quaternion, timestamp, and allocation coverage. Release build is warning-free and all 6,446 tests pass with five intentional skips; retail, architecture, and adversarial reviews are clean.
Co-authored-by: OpenAI Codex <codex@openai.com>
Rewrites src/AcDream.Core/Physics/InterpolationManager.cs from the spec
in docs/research/2026-05-04-l3-port/04-interp-manager.md. Public API
preserved (Vector3-returning AdjustOffset, Enqueue, Clear, IsActive,
Count) so PositionManager + GameWindow callers continue to compile;
internals are full retail spec.
Bug fixes vs prior port (audit 04-interp-manager.md § 7):
#1 progress_quantum accumulates dt (sum of frame deltas), not step
magnitude. Retail line 353140; the prior port's `+= step` made
the secondary stall ratio meaningless.
#3 Far-branch Enqueue (dist > AutonomyBlipDistance = 100m) sets
_failCount = StallFailCountThreshold + 1 = 4, so the next
AdjustOffset call's post-stall check fires an immediate blip-to-
tail snap. Retail line 352944. Prior port silently drifted
toward far targets at catch-up speed instead of teleporting.
#4 Secondary stall test ports the retail formula verbatim:
cumulative / progress_quantum / dt < CREATURE_FAILED_INTERPOLATION_PERCENTAGE.
Audit notes the units are 1/sec (likely Turbine bug or x87 FPU
misread by Binary Ninja) — mirrored byte-for-byte regardless.
#5 Tail-prune is a tail-walking loop, not a single-tail compare.
Multiple consecutive stale tail entries within DesiredDistance
(0.05 m) of the new target collapse together. Retail line 352977.
#6 Cap-eviction at the HEAD when count reaches 20 (already correct
in the prior port; verified).
New API: Enqueue gains an optional `currentBodyPosition` parameter so
the far-branch detection can reference the body when the queue is
empty. Backward-compatible (default null = pre-far-branch behavior).
UseTime collapsed into AdjustOffset's tail (post-stall blip check)
since acdream has no per-tick UseTime call separate from
adjust_offset; identical semantic outcome.
State fields renamed to retail names with sentinel values:
_frameCounter, _progressQuantum, _originalDistance (init = 999999f
sentinel per retail line 0x00555D30 ctor), _failCount.
Tests:
- 17/17 InterpolationManagerTests green.
- New test Enqueue_FarBranch_PrearmsImmediateBlipOnNextAdjustOffset
pins the bug #3 fix: enqueueing 150 m away triggers a same-tick
blip (delta length ≈ 150 m), and the queue clears.
Spec tree: 17 research docs (00–14) under docs/research/2026-05-04-l3-port/.
00-master-plan + 00-port-plan describe the 8-phase rollout. 01-per-tick,
03-up-routing, 04-interp-manager, 05-position-manager-and-partarray,
06-acdream-audit, 14-local-player-audit are the L.3 spec used by this
commit and the M2 follow-up.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Addresses code-quality review findings on commit f43f168:
C-1: Stall detection re-implemented to match retail (acclient lines
353071-353275). Tracks _progressQuantum (sum of step values per window)
+ _distanceAtWindowStart (set at window start). Primary check:
cumulative_progress < MIN_DISTANCE_TO_REACH_POSITION (0.20m absolute).
Secondary check: cumulative_progress / _progressQuantum < 0.30.
Either failing increments fail counter; blip-to-tail at >3 consecutive
fails (already correct).
C-2: Renamed StallFailCountForBlip -> StallFailCountThreshold with
clearer XML doc explaining the > vs >= semantics (blip fires when fail
count EXCEEDS the threshold, i.e. on the 4th consecutive failed window).
I-1: _haveBaselineDistance sentinel prevents first-window false
positive that was triggering spurious fails on every new motion sequence
(old code defaulted _distanceAtWindowStart to 0, making cumulative
progress always negative on frame 5).
I-3: dt <= 0 || NaN guard at AdjustOffset entry prevents NaN
propagation into PhysicsBody.Position.
I-4: Internal field renames for clarity:
_failFrameCounter -> _framesSinceLastStallCheck
_failDistanceLastCheck -> merged into _distanceAtWindowStart
I-5: Added internal Count property + InternalsVisibleTo (via
AssemblyAttribute in .csproj) so Enqueue_DropsOldestWhenAtCap20
actually verifies cap enforcement. Added assertion that head is the
second-enqueued position after overflow.
3 new tests (AdjustOffset_FirstWindow_DoesNotFalseFail,
AdjustOffset_DtZeroOrNegative_ReturnsZero,
Enqueue_AtCap20_HeadIsSecondOriginal), 16 total. All green.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Pure-data class + 13 unit tests.
Ports retail's CPhysicsObj::InterpolateTo (acclient @ 0x005104F0)
and InterpolationManager::adjust_offset (@ 0x00555D30) — FIFO position-
waypoint queue (cap 20) + per-frame catch-up math walking the body
toward the head node at 2 × motion-table-max-speed (clamped, with
7.5 m/s fallback). Reach @ 0.05m. Duplicate-prune @ 0.05m.
Stall detection: every 5 frames; if progress < 30% of expected,
increment fail counter; > 3 fails → blip-to-TAIL (resolved via
decomp dive of UseTime @ 0x00555F20: tail_ is the snap target,
not head_).
Constants verified from binary at named addresses (not guesses):
MAX_INTERPOLATED_VELOCITY_MOD=2.0, MAX_INTERPOLATED_VELOCITY=7.5,
MIN_DISTANCE_TO_REACH_POSITION=0.20, DESIRED_DISTANCE=0.05.
Composed into RemoteMotion in subsequent task; not yet used.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
