Live tp-probe capture refuted the handoff's _datLock-starvation hypothesis:
worker BUILD is fast + uncontended (waited=0ms); the 10-14s 'long transition'
is render-thread APPLY latency, and 'dropped at wrong position' is the
per-frame resolve corrupting the outbound cell frame (lbX zeroed) while the
player sits on an empty world. Design: priority-apply the player's dest
landblock + hold-until-resident behind a retail fade cover (reuse the dormant
TeleportAnimSequencer) + cell-march landblock-id hardening. Foundation-first,
not a hold over slow streaming. Flood-timeslicing + 3D swirl deferred.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Captures the full teleport-flow journey (Slice 1 kept, Slice 2 hold built +
reverted), the root-cause findings (the IsLandblockLoaded key bug + the real
foundation problem: destination doesn't stream fast/complete during teleport,
likely _datLock starvation from the CreateObject flood), the open foundation
issues (#138 slow/incomplete streaming, lost-collision-after-teleport, FPS
leak Work-C, the PortalSpace freeze-vs-run-through question), and the clean
baseline (dd2eb8b). Written as evidence + open questions, not conclusions, so
a fresh session can re-examine the whole approach.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
User-tested: the Slice 2 'hold outdoor until landblock loaded' gate made
EVERY outdoor teleport a ~10 s freeze, because the destination landblock
does NOT load fast during the hold (lbs=0 the whole time — the #138
streaming gap + _datLock starvation from the CreateObject flood). The hold
was band-aiding a broken/slow foundation rather than fixing it, and it never
actually prevented the #145 edge cascade anyway (it force-snapped onto
NO-LANDBLOCK after the timeout regardless).
Reverts ad8c24e..c880973 to the pre-Slice-2 state (00ef47e): outdoor places
immediately again (fast teleports). The genuine bug found along the way —
IsLandblockLoaded queried the wrong key form (& 0xFFFF0000 vs the stored
| 0xFFFF) — is preserved in the history (c880973) and will be re-applied when
we re-introduce a proper hold ON A FIXED FOUNDATION.
Decision (user, 2026-06-21): fix the foundation FIRST — fast/complete
streaming during teleport (#138), the post-teleport lost-collision bug, and
the FPS leak (Work item C) — then revisit the teleport-flow animation. Slice 1
(the pure TeleportAnimSequencer) stays in (dormant, unwired, harmless).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Slice 2 outdoor readiness gate queried IsLandblockLoaded(destCell &
0xFFFF0000) = e.g. 0x7D640000, but streaming stores landblocks under the
EncodeLandblockId form (low 16 = 0xFFFF), e.g. 0x7D64FFFF. The raw
ContainsKey never matched, so the outdoor teleport gate could NEVER flip
Ready and every outdoor arrival ran to the 600-frame (~10 s) timeout and
force-placed. The cascade was still prevented (the timeout force-place lands
cleanly), but the gate did no work — the 10 s freeze the apparatus showed
was this bug, NOT the #138 streaming stall I first suspected.
Root cause found via the apparatus re-test (3-agent investigation
wf_8b67a9d1-35c, all high-confidence) + verified against StreamingRegion.cs:99
(EncodeLandblockId | 0xFFFF), PhysicsEngine.cs:79 (stores as-is),
GameWindow.cs:5530 (queries & 0xFFFF0000).
Fix: IsLandblockLoaded normalizes its arg to the canonical 0xFFFF landblock
key, so the prefix form, any contained cell id, and the dat-id form all
resolve. Added the regression test the original Slice 2 test missed (it had
checked the same 0xFFFF form it added; the real caller passes the 0x..0000
form). Red on the prefix/cell forms before the fix, green after. 9/9.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Code-review follow-up: the exit-sound/login-complete events are emitted
inline at their transitions, so the _exitSoundPending/_loginCompletePending
fields were set-then-cleared dead code — removed. Added a comment explaining
why TunnelContinue's min-advance is gated on worldReady. No behavior change;
29/29 sequencer tests still green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
AP-49: TeleportAnimSequencer.ComputeFadeAlpha uses smoothstep in place of
retail's unrecovered 1024-entry GetAnimLevel lookup table
(gmSmartBoxUI::UseTime 0x004d6e30). Retire when the table contents are
extracted via cdb (spec §8).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
EnterTunnel fires on the first Tick after Begin for Portal/Login/Death kinds
(which enter directly at Tunnel). Already implemented in Task 1.2 via
_enterTunnelPending = _state == TeleportAnimState.Tunnel.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
TunnelContinue exit gate: minMet requires worldReady (min-continue hold);
maxForce fires unconditionally at MaxContinue (safety-net fallback when
world never loads). This matches spec §3.4.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Bite-sized TDD plan for the retail teleport flow. Slice 1: pure
TeleportAnimSequencer (7-state TAS) + golden-timing tests. Slice 2: the
#145 readiness-gate fix (IsLandblockLoaded + Decide outdoorReady axis +
apparatus probe) — ships independently of the visuals. Slice 3:
TeleportFlowController (delegate-injected, unit-tested) + TeleportFadeOverlay
+ portal wiring (PlaceTeleportArrival split: place vs InWorld so the input
lock persists the whole animation). Slice 4: one yaw-freeze + portal sounds
via the EnumIDMap chain. Slice 5: de-dup login readiness onto Decide, route
login/death through the controller, logout (Shift+Esc) + 0xF653 + disconnect,
remove dead _teleportArrival plumbing. Slice 6 (literal 3D swirl) is a
follow-up plan gated on a cdb asset trace.
Drafted via two research + drafting workflows; slices 3-5 redrafted as one
cohesive unit against a pinned controller API after the first parallel pass
produced cross-slice inconsistencies (missing controller task, triplicated
yaw-freeze, a fabricated PlayerMovementController.Update signature). All
load-bearing signatures personally verified against the tree.
Plan: docs/superpowers/plans/2026-06-21-retail-teleport-flow.md
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
3rd live session found the carried-anchor fix is incomplete: the cascade recurs
when a teleport arrives onto a NOT-YET-STREAMED landblock near an edge (0xC98C
arrival at local Y=190.3, NO-LANDBLOCK -> marches 0x8C->0xFE, wire localY=-21684,
ACE rejects). Streamed-arrival case IS fixed (verified ~10 landblocks). Same root
as the Z free-fall (#135/#138 placed-but-unstreamed gap). Prior 'gate passed' was
premature. Needs apparatus (anchor/guard diagnostic at the crossing) before a fix.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adversarial review flagged a possible CellPosition staleness on indoor->outdoor
transition. Verified against source: false positive (SnapToCell isn't called on
building entry; the body is world-space so the delta is frame-invariant; the anchor
disengages indoors). Added a test proving CellPosition tracks the outdoor cell under
the world position across a multi-cell, cross-landblock walk and stays canonical.
Core cell-sync 6/6.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
SetPosition zeros the body velocity but the motion interpreter kept the
PRE-teleport ForwardCommand (RunForward), so the next Update() rebuilt that
run vector via get_state_velocity and the player sprinted off in the old
direction on arrival. DoMotion(Ready) makes the player arrive at rest. Not a
cascade fix anymore (Slice 3 closed that) — the last user-visible bit of #145.
All suites green: Core 1529 / App 480 / UI 425 / Net 313.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The outdoor membership pick derived the landblock origin from the terrain
registry, which returns (0,0) for an UNSTREAMED neighbour — so a fresh far-town
teleport at a landblock edge marched the cell id one block per physics tick
(the cascade; the 17410 ACE rejects is its wire artifact).
Fix: thread the CARRIED cell-relative frame anchor (body.Position -
body.CellPosition.Frame.Origin) into the pick via SpherePath.CarriedBlockOrigin.
That anchor IS the true landblock world origin, correct even for an unstreamed
neighbour, so the pick re-derives the SAME (consistent) cell and never marches.
- CellTransit.FindCellSet/BuildCellSetAndPickContaining: Vector3? carriedBlockOrigin
(null default = legacy TryGetTerrainOrigin → every existing caller/test untouched).
- PhysicsEngine.ResolveWithTransition: set the anchor from a SEEDED OUTDOOR body
whose carried landblock matches the resolve cell (else null → legacy).
- PlayerMovementController.SetPosition: 3-arg overload seeds CellPosition from the
wire's (cell, local) via SnapToCell; 2-arg delegates with cellLocal=pos (anchor
(0,0,0) == legacy → zero test churn).
- GameWindow.CellLocalForSeed: the placement seam (_liveCenter used ONCE here to
derive the cell-local; physics carries it forward without _liveCenter).
Regression: TeleportFarTownRunawayTests (south + east edge, unstreamed neighbour).
Core 1529 / App 480, zero regressions.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The wire local is LANDBLOCK-relative [0,192); the cell low word = floor(local/24),
so a consistent (cell,local) pair must keep them in lockstep. Two retail-faithful
corrections vs the first pass:
- SnapToCell canonicalizes the OUTDOOR seed via AdjustToOutside (retail
SetPositionInternal/adjust_to_outside @0x00504A40 — the #107 'never trust a
server (cell,pos) pair' protection). Indoor seeds stay verbatim (BSP-validated).
- SyncCellPositionDelta calls AdjustToOutside on EVERY delta, not just on 192 m
crossings, so intra-landblock 24 m cell-index changes track (needed by Slice 3
membership). Idempotent within a cell.
Tests rewritten to verify both (the earlier test paired an inconsistent cell+local).
Core 1527 passed / 0 failed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add CellPosition (retail Position type) alongside the world Vector3 Position.
The Position setter mirrors each world delta into the cell-local origin and
calls AdjustToOutside only when the local coord crosses a landblock boundary
([0,192) on X or Y), so the within-block cell id is preserved from the wire
seed. SnapToCell seeds both positions from the wire's (cell, local) pair
verbatim — no streaming center involved. Unseeded bodies (ObjCellId==0) and
indoor cells are no-ops in the delta path. UpdatePhysicsInternal's existing
`Position +=` desugars through the new setter automatically; no call sites
changed. 4 new unit tests; full Core suite 1526 passed / 0 failed / 2 skipped.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The new value type collided with DatReaderWriter.Types.Frame (used in
physics-adjacent code like ShadowShapeBuilder), which the structural fix
(per-file using-aliases across 6 files) would have re-incurred in every
later physics slice. Renamed the TYPE to CellFrame; the Position.Frame
MEMBER keeps retail's name. Restored the 5 alias-only files to their
pre-Slice-1 state; synced spec + plan. Core 1522 passed / 0 failed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Introduces the two value types (Frame, Position) that represent retail's
cell-relative position pair (acclient.h:30647/30658). Types are unused
by consumers yet — zero behavior change. Also ports LandDefs::get_block_offset
(pc:69189, @0x0043e630): world-meter offset between two named landblock ids,
the ONLY cross-cell translation primitive in retail physics. Conformance tests:
same-landblock→Zero, south-neighbour→(0,-192,0) (the exact #145 cascade cell),
east-neighbour→(+192,0,0), diagonal→(+192,+192,0). 4/4 pass; full Core suite
1522 passed / 0 failed. DatFrame alias added to 4 files that had using
DatReaderWriter.Types + using AcDream.Core.Physics in scope simultaneously.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Approved design for the cell-relative physics frame port: retire
_liveCenter from physics (render-only), carry Position{ObjCellId,
Frame{local in [0,192), quat}} as the source of truth, port
get_block_offset (verified 0x0043e630 nets to delta-landblock * 192 m),
translate inter-tick collision state per-cell like retail (0x0050a592).
7-slice parallel-frame migration; conformance via transform-on-read (no
fixture re-capture).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Workflow wf_87607d15-c43 (4 research streams + synthesis + 3 adversarial verifiers, HIGH confidence) confirmed the far-town runaway is a cell-membership label cascade from a discarded TryGetTerrainOrigin bool (CellTransit.cs:736 -> (0,0) origin for unstreamed neighbors), not a free-fall; 17410 is a wire artifact. User chose the architectural fix: port retail's cell-relative Position + retire _liveCenter from physics. Handoff doc carries the verified mechanism, the retail port table (decomp addresses), acdream divergence sites, apparatus (desync-capture.jsonl + probes + harness template), and the brainstorming-gate requirement.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Captured: teleport to far town (201,91) places correctly via the #145 verbatim path, then the per-frame resolve marches membership one landblock south/frame (un-rebased local position) until ACE rejects the inconsistent (cell, local) pair. #138 re-hydrate exonerated. Root cause under multi-agent research (acdream code + retail decomp oracle + capture).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Records the confirmed root cause (ACE never clears KnownObjects on a
teleport so it won't re-send known objects; retail/holtburger keep the
client object table and re-render from it) and the two-part fix
(re-hydrate from _lastSpawnByGuid; pending-bucket persistent rescue).
Corrects the 2026-06-21 handoff: ClientObjectTable is the inventory data
model with no world position/Setup and cannot rebuild a render entity;
the real retained world-object table is GameWindow._lastSpawnByGuid.
Status: FIX SHIPPED, pending user visual gate.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
GpuWorldState.RemoveLandblock rescued persistent entities (the player)
only from the _loaded list, silently dropping one sitting in the
_pendingByLandblock bucket. The player is re-injected via AppendLiveEntity
every frame; right after a teleport its destination landblock has not
streamed in yet, so the player lands in the pending bucket — and if that
landblock is then unloaded during the streaming churn, the persistent
entry was dropped, violating the "persistent therefore survives unload"
contract. Leading candidate for the #138 "own avatar vanishes after a
couple round-trips" symptom (cumulative; needs user visual confirm).
Fix: scan the pending bucket for persistent guids and rescue them too,
so DrainRescued re-parks them at the next valid landblock. Provable
correctness fix with a deterministic test (rescue-from-pending plus a
negative for non-persistent). Core tests green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Doors/NPCs/portals vanished after a portal OUT of the 0x0007 dungeon back
to Holtburg. Root cause confirmed via ACE + holtburger cross-reference:
the dungeon collapse drops a landblock's render entities for FPS, and ACE
will NOT re-broadcast objects whose guid is still in its per-player
KnownObjects set (never cleared on a normal teleport — ACE relies on the
client retaining its object table and culling stale objects itself). So
nothing restored them on the way back.
Retail-faithful fix: a real client keeps its weenie_object_table and
re-renders the world from it (holtburger keeps the table across a
teleport; only suspends physics bodies). acdream's _lastSpawnByGuid (the
parsed CreateObject records — position + Setup + appearance) IS that
table and survives the collapse (the collapse path never calls
RemoveLiveEntityByServerGuid, the only thing that prunes it). On landblock
(re)load, replay OnLiveEntitySpawnedLocked for retained spawns whose
render entity is absent — independent of any ACE re-send.
- LandblockEntityRehydrator: pure selection (landblock match; skip
already-present, the player, and mesh-less spawns), unit-tested (7).
- StreamingController: onLandblockLoaded callback after AddLandblock
(Loaded = dungeon-exit expand) and AddEntitiesToExistingLandblock
(Promoted = Far->Near).
- GameWindow.RehydrateServerEntitiesForLandblock: present-gate keys on
GpuWorldState (NOT _entitiesByServerGuid, which holds collapse
orphans), replay under _datLock; the replay's own
RemoveLiveEntityByServerGuid de-dup scrubs the orphan state.
Corrects the handoff: ClientObjectTable is inventory-only (no world
position/Setup) and cannot rebuild a render entity; _lastSpawnByGuid is
the world-object table. Register row AP-48 (no retail 25s visibility
cull). dotnet build + 1518 Core tests green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Full orientation for a fresh session to fix#138: confirmed root (server
objects unloaded on teleport-IN, not restored on return; ACE re-broadcast
unreliable), DO-NOT-RETRY table (cache + render-cull eliminated), the
ClientObjectTable re-hydrate fix direction with file:line pointers, the
launch/probe/account setup, and the gotchas (re-broadcast latency, stale
sessions, don't-kill-clients, entity.Id != ServerGuid).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Deep dive (this session) eliminated the 2026-06-20 hypotheses for #138
(server objects + own avatar not showing after a teleport-out):
- NOT the Tier-1 classification cache: re-created live entities get a fresh
monotonic Id (_liveEntityIdCounter++), so the cache (keyed on Id) is always
a miss for them. (Side-finding: the cache has a real demote-vs-unload
invalidation asymmetry — RemoveLandblock doesn't fire _onLandblockUnloaded
while RemoveEntitiesFromLandblock does — but it's NOT the #138 cause.)
- The render path is fine when entities are present (login: [dyn] dyn=54
drawn=33; the dynamics partition + DrawDynamicsLast draw them).
- The actual cause: re-delivery is unreliable. notan/+Je walk-around run after
teleport-out: live:spawn doors=0, [ent]+ door appends=0, [ent-flat] server=1
— the server delivered ZERO Holtburg objects on return; they never reach
acdream. acdream unloads them on teleport-IN (the collapse) and nothing
restores them; ACE doesn't reliably re-broadcast. "Other clients see +Je"
confirms it's acdream's local world, not server state.
Fix direction (next session): re-hydrate GpuWorldState from the retained
ClientObjectTable on AddLandblock instead of depending on an ACE re-broadcast
(or treat in-range server objects as persistent across the collapse). Entity-
lifecycle/protocol change, best started fresh.
Diagnostic scaffolding (probes + the unrelated cache-asymmetry fix) reverted;
tree is back at the green #145 state (a15bd3b).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Portals only worked once per session: teleporting OUT of a dungeon
mis-rooted the player into the SOURCE dungeon's coordinate frame, so every
move was sent dungeon-framed and ACE rejected it ("failed transition") —
the player couldn't move, never reached a portal, and the world wouldn't
re-render (only skybox).
Root cause: acdream's streaming-relative frame recenters on teleport, but
resident physics landblocks keep their load-time world-offset. After
recentering onto the outdoor destination, the collapsed source dungeon
(offset 0,0 as the prior center) and the destination (offset 0,0 as the
new center) overlap, and the Z-agnostic outdoor cell-snap returns the
dungeon for both the arrival placement and every per-frame resolve.
Fix (server-authoritative teleport placement):
- Drop the stale source center landblock from physics at the teleport
recenter (GameWindow.OnLivePositionUpdated) so the resolve falls through
to the server position (Resolve NO-LANDBLOCK verbatim) until the
destination streams in.
- Place outdoor teleports immediately (TeleportArrivalRules) — holding is
futile because streaming does not progress during a PortalSpace hold.
- Clear a dangling CellGraph.CurrCell when its landblock is removed
(PhysicsEngine.RemoveLandblock) — otherwise the dungeon-streaming gate
keeps streaming collapsed onto the gone dungeon (only skybox renders).
Keeps DungeonStreamingGate (gate suppression during the hold). Indoor
(dungeon-entry) placement is unchanged (cell-keyed, IsSpawnCellReady).
User-verified: in->out->re-enter works repeatedly, no ACE errors, world
renders. Remaining facets (server objects + own avatar not rendering after
a teleport-out) are entity render/lifecycle — split to #138.
Registers AP-36 + AD-2 updated. New: DungeonStreamingGate (+4 tests),
TeleportArrivalRules (+4 tests). Build + 2727 tests green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Indoor lighting DONE this session (#142/#143 closed). Per user: do NOT merge to
main; start fresh next session. NEXT-session order set in the handoff:
1. #145 — portals only work once per session (run in/out/re-enter repeatedly).
Machinery: 0xF751 PlayerTeleport -> PortalSpace -> TeleportArrivalController ->
streaming collapse/expand. Likely overlaps #138. Instrument a 2nd teleport first.
2. #144 — dungeon interiors still too dim vs retail (cdb side-by-side first).
Handoff: docs/research/2026-06-20-indoor-lighting-done-next-portal-reuse-handoff.md
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
#142 + #143 resolved this session — interiors + the meeting-hall portal now match
retail (user-confirmed). The #142 diagnosed cause (per-frame sun/ambient regime)
was a red herring; the real bug was the EnvCellRenderer landblock-key lookup
(0d8b827) that starved every interior wall of point lights. #143's portal light
rides the weenie-light path + the dynamic D3D 1/d attenuation (57c2ab7).
#144: dungeons improved (torch cells light up now) but torch-sparse stretches +
overall brightness still trail retail. Needs a side-by-side cdb capture of
retail's dungeon (active lights + ambient) — candidates: per-vertex bake under-lit
on low-poly walls, 0.2 sealed ambient too dark, or retail leans harder on dynamics.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The portal swirl's magenta light (and the viewer fill) read as a tight,
concentrated pool vs retail's soft, room-wide tint. Cause: acdream applied its
STATIC dat-bake falloff (1/d^3 distance-cube + range x1.3) to ALL point lights,
including dynamic ones. Retail draws dynamic lights through the D3D hardware
path (config_hardware_light 0x0059ad30): a point light gets Attenuation1=1 =>
att = 1/d (inverse-linear), plain Lambert, range x1.5 (rangeAdjust 0x00820cc4).
Split the two paths by a per-light IsDynamic flag:
- LightSource.IsDynamic; packed into GlobalLight.coneAngleEtc.y (binding=4).
- LightInfoLoader.Load(isDynamic) => range x1.5 + flag (server-object/portal
lights via the live spawn path); dat-static lights keep x1.3 (default).
- Viewer fill + weenie/portal lights = dynamic; dat torches = static.
- mesh_modern.vert pointContribution: dynamic branch = 1/d att, plain Lambert,
hard cutoff, no per-light cap (D3D accumulates then saturates via the existing
min(pointAcc,1)); static branch = the unchanged wrap/norm bake.
This is the portal half of #143 (the magenta light itself now registers + reaches
the walls via the prior weenie-light + landblock-key fix). Refines AP-35: point
lights now split static-bake (1/d^3) vs dynamic-hardware (1/d) by path.
Verified: portal light now range=9 (6x1.5), magenta spreads softly; shader
compiles clean; static torches unchanged (range 5.2/6.5/7.8). User-confirmed the
portal matches retail and the torch-lit interior did not over-brighten.
Core lighting 44/44, App 476 green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The whole "indoor interiors read dark/flat vs retail" saga was ONE root-cause
bug: EnvCellRenderer.GetCellLightSet derived the landblock key as
`cellId & 0xFFFF0000` (0xXXYY0000), but landblocks are keyed by the streaming
id 0xXXYYFFFF. The lookup missed for EVERY cell, so SelectForObject never ran
and every EnvCell wall received ZERO point lights — torches, lanterns, the
viewer light, all of it. Confirmed by a [cell-light] probe: inBounds=False
selected=0 across 1M+ cell draws; after the fix inBounds=True selected=3-4.
User-confirmed the interiors now look like retail.
Three faithful additions that were blocked by the key bug (and only show now):
- Viewer light (LightManager.UpdateViewerLight): retail's SmartBox::set_viewer
(0x00452c40) adds a white fill light at the player every frame via
add_dynamic_light — the dominant interior fill (no sun indoors). acdream had
NO dynamic lights at all. Params from the cdb capture: intensity 2.25,
falloff 10, white, offset (0,0,2). Indoor-only via the AP-43 gate.
- Weenie fixture lights (OnLiveEntitySpawnedLocked): server-spawned lanterns/
braziers carry Setup.Lights but the dat-static registration never saw
CreateObject entities. Register on spawn; unregister on despawn
(UnregisterOwner made unconditional). Register row AP-44.
- IndoorObjectReceivesTorches now excludes the 0xFFFF landblock marker (it is
not an EnvCell) — fixes WbDrawDispatcherIndoorFlagTests.LandblockId_OutdoorFlag0
(a #142 verification miss).
Divergence register: AP-44 (weenie light spawn-position, no movement tracking),
AP-47 (acdream's 128-light/camera-independent cap keeps interiors always-lit vs
retail's 40-nearest-to-player budget that pops in on approach — intentional,
user-preferred).
Investigation: docs/research/2026-06-20-indoor-torch-lantern-lighting-investigation.md
Core 1505 / App 476 green. Visual gate: user-confirmed "looks like retail now."
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Standing inside (or looking into) a windowed building like the Agent of
Arcanum, interior objects (furniture, NPCs, the player) were lit by the
directional sun because acdream's sun gate was per-FRAME (keyed on the
player being in a sealed cell), not per-DRAW as retail does it.
Retail's PView::DrawCells (0x005a4840) runs two stages per frame:
outdoor stage → useSunlightSet(1) (0x005a485a): sun ON
interior stage → useSunlightSet(0) (0x005a49f3): sun OFF
DrawMeshInternal (0x0059f398) then calls minimize_object_lighting only
when useSunlight==0, so indoor objects ALWAYS skip the sun regardless of
whether the player's cell is windowed or sealed.
Fix: add a per-instance uint SSBO (binding=6 instanceIndoor[]) whose value
is IndoorObjectReceivesTorches(ParentCellId) — the same predicate AP-43
already uses for the torch gate. In mesh_modern.vert, nest the sun loop
inside an additional `if (instanceIndoor[instanceIndex] == 0u)` check
inside the existing `if (uLightingMode == 0)` block. Indoor objects get
torches (unchanged) but now skip the sun; outdoor objects keep the sun and
still get no torches. The ambient regime (UpdateSunFromSky: 0.2 sealed /
sky otherwise) is untouched — it was already correct.
Mechanically: _currentEntityIndoor set once per entity in
ComputeEntityLightSet; appended to InstanceGroup.IndoorFlags in
AppendCurrentLightSet; grown/packed/uploaded in the same cursor loop as
_clipSlotData and _lightSetData; deleted in Dispose. Mode-1 draws
(EnvCellRenderer) never read binding=6 — the sun loop is inside the
uLightingMode==0 uniform-control-flow branch.
AP-43 divergence register updated: the sun half is now per-draw (no
longer a residual). Residual narrowed to the unaudited ebp_2 test in
CellManager::ChangePosition (no observed impact).
Tests: WbDrawDispatcherIndoorFlagTests pins IndoorObjectReceivesTorches
for the spec §5 representative ids: 0xA9B40172 (Agent of Arcanum EnvCell)
→ 1; 0xA9B40031 (land sub-cell) → 0; 0xA9B4FFFF (landblock) → 0; null
(outdoor shell) → 0; plus the boundary cases 0x0100/0x00FF.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Decomp + in-game probe (agent-arcanum-probe.log) show the ambient regime
is already retail-faithful (CellManager::ChangePosition 0x004559B0); only
the sun is wrong — acdream gates it per-FRAME, retail per-STAGE
(useSunlight 0x0054d450). Fix = a per-instance indoor flag (reusing
IndoorObjectReceivesTorches, the AP-43 predicate) gating the sun off for
indoor mode-0 objects. No second ambient, UpdateSunFromSky + EnvCellRenderer
unchanged. User pre-approved.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
