The sky + weather (rain cylinder) are retail's LScape — 'the outside seen through the exit portal.' Retail PView::DrawCells (pseudo_c:432709) draws LScape clipped to the OutsideView when outside_view.view_count>0, then does a conditional Z-buffer-ONLY clear (432731) before the indoor cells. acdream now does the same:
- sky.vert writes gl_ClipDistance against the SAME binding=2 TerrainClip UBO the terrain reads. The OutsideView planes are screen-space (NDC) half-spaces encoded as clip-space planes (nx,ny,0,dw); the test dot(plane,gl_Position)>=0 reduces after perspective divide to nx*ndcX+ny*ndcY+dw>=0 — projection-INDEPENDENT — so the same plane set clips the sky EXACTLY despite its separate dome projection. count==0 (outdoor) → all distances +1 → full-screen, bit-identical. Lighting/fog math untouched.
- GameWindow: relocated the sky pre-scene + weather post-scene draws to their retail LScape positions, each in a local 8-plane clip bracket so sky.vert confines them to the doorway indoors / full-screen outdoors. Added the conditional doorway depth-ONLY Z-clear (no color → no blue hole), scissored to the OutsideView AABB. drawSkyThisFrame = seen_outside policy AND (outdoor OR exit-portal-in-view) — a sealed interior with no exit portal in view draws no sky (kills the full-screen-sky interim regression). Sky pre/post particle passes (particle.vert has no gl_ClipDistance) scissored to the doorway bbox.
- ClipFrameAssembly gains HasOutsideView + OutsideViewNdcAabb (the doorway NDC AABB, computed for BOTH Planes and Scissor terrain modes — unlike TerrainScissorNdcAabb which is Scissor-only).
- The pre-login goto SkipWorldGeometry moved BELOW the sky draw so the live sky still renders during the EnterWorld handshake (clipAssembly is null/no-clip pre-login → full-screen).
Build green; App tests 160/160. Stage 4 tests + verify-annotations follow.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
ComputeVisibilityFromRoot(null, …) now returns null (outdoor root) instead of
calling FindCameraCell(fallbackPos). Retail CellManager::ChangePosition
(0x004559B0) reads the transition-owned curr_cell — it does NOT re-derive from
a static position. W2a guarantees CurrCell is set from the first tick, so the
AABB fallback is dead. Deleted: FindCameraCell (389–446), _lastCameraCell,
_cellSwitchGraceFrames, CellSwitchGraceFrameCount. GetVisibleCells retains a
brute-force AABB scan for test-compat; ComputeVisibility stays for the same
reason. Updated 3 null-root tests in CellVisibilityFromRootTests to assert the
new null-returns-null behavior.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Wire the BFS visibility root to DataCache.CellGraph.CurrCell (the physics
membership answer written in W2 Task 1) rather than resolving independently
from a position via FindCameraCell. Closes the render/physics disagreement
that causes the "world from below" spawn-in flicker.
Changes:
- CellVisibility.GetVisibleCells: extracted BFS body into new private
GetVisibleCellsFromRoot(LoadedCell root, Vector3 cameraPos); existing
GetVisibleCells delegates to it after FindCameraCell (behavior unchanged).
- CellVisibility.ComputeVisibilityFromRoot(LoadedCell? root, Vector3 fallbackPos):
new public entry point; when root is null falls through to ComputeVisibility
(exact today's behavior), otherwise sets _lastCameraCell = root and delegates
to GetVisibleCellsFromRoot — cannot regress below baseline.
- GameWindow (line 7156): replaced ComputeVisibility(visRootPos) with
ComputeVisibilityFromRoot(physicsRoot, visRootPos) where physicsRoot is
resolved from _physicsEngine.DataCache.CellGraph.CurrCell via TryGetCell.
physicsRoot is null whenever CurrCell is null or its id is not yet in the
render registry, so the fallback fires until the cell loads.
- 6 new tests in CellVisibilityFromRootTests: null-root fallback equivalence
(3 cases), registered root → CameraCell == root (3 cases). All 160 App.Tests
pass, 0 regressions.
Visual verification PENDING — behavior change; do not claim it works visually.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A week on the indoor render (Phase U.4 → U.4c → 2026-05-31) fixed the flap but
produced NO shippable progress: walls/ceiling don't seal, outdoor terrain is
visible from inside (#78), the enclosure reads grey/transparent. Root cause is
ARCHITECTURAL, not a bug.
Evidence this session (direct, via the new [shell] probe + screenshots) RULED OUT
every subsystem except the gating architecture: the interior cell shells render
fine (geometry/texture/opaque/depth all correct, zh=0 tr=0); the visibility
traversal computes correct sets + non-empty portal clips; cull mode is fine; the
camera/eye thread was a detour. The residual is that OUTDOOR geometry is not gated
to portal openings when indoors, and acdream enforces visibility THREE inconsistent
ways (TerrainClipMode / per-cell shell clip / entity ParentCellId filter with an
outdoor-stab bypass) instead of retail's ONE PView gate.
This commit is the reset handoff + documentation, not a code fix:
- docs/research/2026-05-31-render-architecture-reset-handoff.md — canonical: honest
state, evidence ledger (ruled-out / do-not-repeat), the mapped 3-gate patchwork,
the retail PView target (one traversal → one gate for ALL geometry), the reset
mission, and a copy-paste pickup prompt.
- docs/architecture/acdream-architecture.md — new "Render Pipeline" SSOT section
(current divergence + unified-PView target + the one rule: compute visibility
once, enforce it once). (Doc has pre-existing corruption below this section —
flagged for separate cleanup.)
- Apparatus: ACDREAM_PROBE_SHELL → [shell] (EnvCellRenderer per-cell prepared/drawn
geometry + flags) added to RenderingDiagnostics + EnvCellRenderer. Throwaway.
- docs/superpowers/specs/2026-05-31-camera-collision-indoor-engagement-design.md —
spec for e099b4c (camera collision; now parked as orthogonal to the seam).
Next session: STOP point-fixing; do the architecture reset to a single PView gate.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Root cause (b): ShadowObjectRegistry.GetNearbyObjects (line 480) returns early
when primaryCellId is an indoor cell, skipping the outdoor radial sweep that
contains the landblock-baked cottage exterior-shell GfxObj. The issue-#98 fix
that prevents the player's head sphere from being capped by the cottage floor
also prevents the IsViewer camera sweep from finding the exterior building shell.
Result: camera passes through exterior walls unimpeded, driving the residual
transparent-walls symptom after the U.4c flap fix.
Evidence: live capture shows eyeInRoot=n ~90% of frames, eye-player distance
3.43m (full chase, no pull-in). RED test deterministically reproduces: synthetic
indoor cell (0xA9B40175) + exterior GfxObj registered at cellScope=0; probe
SweepEye returns pulledIn=0.0000m (full eye distance Y=5.0, wall at Y=4.0).
Fix design: exempt IsViewer from the indoor-primary early-return gate in
GetNearbyObjects — retail's find_obj_collisions (named-retail :308918) has no
indoor/outdoor cell gate; the acdream fix is correct only for IsPlayer.
Apparatus committed:
- tests/AcDream.App.Tests/Rendering/CameraCollisionIndoorTests.cs (RED test)
- docs/research/2026-05-31-camera-collision-indoor-diagnosis.md (findings + design)
- PhysicsCameraCollisionProbe.cs [flap-sweep] diagnostic retained (U.4c spike)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The visibility root + portal-side test now use the PLAYER position (visRootPos) in
player mode instead of the camera EYE; the eye still drives the per-frame projection
(envCellViewProj). Live ACDREAM_PROBE_FLAP evidence: the flap was the 3rd-person eye
drifting out of the player's cell -> FindCameraCell returning the STALE cell for its
grace frames -> the doorway portal culled as behind-the-eye -> exit cell + terrain +
shells dropped (res=Grace eyeInRoot=n terrain=Skip on every flap frame). Retail's
CellManager::ChangePosition (0x004559B0) tracks curr_cell by the player; acdream
already roots lighting at the player (GameWindow:7152) for the same chase-cam reason
— visibility was the lone holdout on the eye. Removed the earlier synthetic builder
flap test, which modeled a disproven (side-test) hypothesis; the fix is integration-
level, validated by the visual gate + [flap] probe. App tests 151/151.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The [flap] line now reports, per root portal, the actual projected NDC vertices and
the Intersect-against-FullScreen result count (clip=N), so a portal that PROJECTS
(proj>=3) but still fails to ADD its neighbour (vis stays low) shows WHY: clip=0 with
ndc inside [-1,1] = winding/self-intersection degeneracy; clip=0 with ndc outside
[-1,1] = genuinely off-screen; the ndc coords expose a near-plane bowtie. Pins the
exact clip-region failure before the root-cause fix.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Per-frame (not cell-change-throttled, so it catches the flicker at a stable root):
[flap] line from the builder — root cell's per-portal side-test D + traverse/cull +
NDC projection, plus OutsideView poly count + visible-cell count; localEye exposes
when the eye has crossed an interior portal plane. Paired [flap-cam] line from the
draw site — FindCameraCell resolution branch (CameraCellResolution enum, new),
eyeInRoot AABB flag (stale-root signal), eye + player worldpos, and the frame's
TerrainMode/OutdoorVisible outcome. Disambiguates side-cull vs empty-projection vs
stale-root. Inert when off (gated). Throwaway apparatus to converge the flap fix.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
VisibleCells (full ids) + SeenOutside, populated at the EnvCell-build site from
envCell.VisibleCells + envCell.Flags. Mirrors retail CEnvCell.stab_list /
seen_outside (acclient.h ~30925). Data already in-process; render path no longer
drops it. Consumed by the builder in U.4c-3.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Second self-contained-GL-state fix (after uViewProjection): EnvCellRenderer.Render set
BlendFunc per-batch but never the BLEND enable or DepthMask. The opaque shell pass —
drawn after terrain (which sets neither) and after particles / last frame's transparent
pass — inherited whatever left GL_BLEND enabled, making opaque walls composite their
sub-1.0-alpha textures against the bluish clear color (terrain Skip'd indoors) →
"transparent walls / only background," flickering with per-frame ordering. Mirror the
working WbDrawDispatcher: Disable(Blend)+DepthMask(true) opaque, Enable(Blend)+
DepthMask(false) transparent, restore opaque defaults after the draw loop.
Does NOT address the threshold "flap" (OutsideView instability from the per-frame
view-dependent BFS) — that is a distinct, deeper root cause tracked separately.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Root cause of the indoor cell-shell SEAM flicker ("transparent walls, oscillating
when moving"): EnvCellRenderer.Render never set uViewProjection — it inherited
WbDrawDispatcher's. But the opaque shell pass draws BEFORE the dispatcher's Draw
(GameWindow ~7411 vs ~7418, the only other setter), so opaque shells rendered with
the PREVIOUS frame's matrix — a stale gl_Position against this frame's clip planes,
yielding pose-dependent clipping that's worst while moving. Make Render self-contained:
stash the view-projection in PrepareRenderBatches and upload it in Render (same matrix
the portal clip planes use). Same self-contained-GL-state precedent as the 2026-05-28
cull-state fix in this file.
Visual re-test confirms this removes the wall-seam flicker. A separate residual
("some houses show only background on interior walls; some flicker remains") is a
distinct root cause, under investigation — NOT this matrix bug.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Code review flagged the gate-critical per-instance slot resolution as untested.
Add RED→GREEN cases (live=unclipped slot 0, cell-static→cell slot, non-visible→cull,
outdoor-stab→OutsideView/cull, routing-inactive→all slot 0). Note the full-cell-id-space
invariant at ResolveEntitySlot; fix a stale RenderInsideOut comment in EnvCellRenderer.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Wire the portal-visibility result through the clip pipeline: build a per-frame
ClipFrame (slot 0 no-clip, slot 1 OutsideView, slot 2..N per visible cell) +
cellIdToSlot from PortalVisibilityBuilder; call the (previously dormant)
EnvCellRenderer.Render for cell shells inside the clip bracket; assign per-instance
clip slots in WbDrawDispatcher (live-dynamic unclipped per retail, cell statics to
their cell slot, outdoor scenery to OutsideView, non-visible culled); gate/scissor/
skip terrain per OutsideView (empty ⇒ no terrain — the bleed fix). Emit ACDREAM_PROBE_VIS.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Code review caught a portability hazard: GL_CLIP_DISTANCE0..7 was enabled globally
at init, but sky/particle/ui/debug vertex shaders don't write gl_ClipDistance —
undefined behavior that could clip them away on some drivers (benign on the dev
driver, which is why the offline check passed). Bracket the enable/disable around
only the terrain+entity (mesh_modern/terrain_modern) draws; sky/particles/UI/debug
render with clipping off. U.4's EnvCellRenderer.Render belongs inside the bracket.
Also: ClipFrame is long-lived (??= NoClip()), so Dispose now deletes its GL buffers;
fix the stale per-frame-transient comments.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds the GPU mechanism to clip drawing to a per-cell screen-space convex
region via gl_ClipDistance, consumed by the mesh + terrain vertex shaders.
This is the MECHANISM only — every instance defaults to slot 0 (no-clip /
pass-all) and terrain to count 0, so the running game renders IDENTICALLY to
pre-U.3 (verified: offline launch compiles both shaders and reaches steady
state; no GL errors). U.4 populates real clip data from portal visibility.
Binding contract (define once, both sides obey):
- mesh_modern.vert: SSBO binding=2 CellClip[] (shared per-frame regions, slot 0
reserved no-clip) + SSBO binding=3 uint[] per-instance slot, indexed by the
IDENTICAL gl_BaseInstanceARB+gl_InstanceID used for binding=0. binding=0/1
untouched.
- terrain_modern.vert: UBO binding=2 TerrainClip { int count; vec4 planes[8]; }
for the single OutsideView region (UBO namespace; SceneLighting is UBO
binding=1, so binding=2 is free and does not collide with the mesh SSBO
binding=2). count 0 = ungated.
- Both redeclare out gl_PerVertex { vec4 gl_Position; float gl_ClipDistance[8]; }
and set unused planes (i >= count) to +1.0 so they pass everything.
CellClip std430 layout (144 bytes/slot): count@0, 3 pad uints@4/8/12,
planes[8]@16 (vec4 stride 16). Terrain UBO std140: count@0 (padded to 16),
planes[8]@16 → 144 bytes. Verified by ClipFrameLayoutTests (8 new tests).
Pieces:
- ClipFrame: per-frame container + uploader for the SHARED clip data (binding=2
SSBO + terrain UBO). NoClip() = slot 0 + terrain count 0. AppendSlot /
SetTerrainClip pack std430/std140 bytes for U.4. UploadShared binds both.
- WbDrawDispatcher + EnvCellRenderer: each owns its binding=3 zero buffer
(all-zeros sized to its instance count → slot 0), re-binds binding=2 from the
shared ClipFrame id (or an internal no-clip fallback if unwired) before MDI.
gl_ClipDistance is per-vertex, so the single glMultiDrawElementsIndirect per
group is preserved — no draw splitting.
- TerrainModernRenderer: binds the terrain clip UBO (shared or no-clip fallback)
before its draw.
- GameWindow: glEnable(GL_CLIP_DISTANCE0..7) once at init (unused planes pass-all
so always-on avoids per-draw thrash); per frame builds ClipFrame.NoClip(),
UploadShared, and hands the buffer ids to the three renderers (tiny diff; U.4
swaps NoClip() for the real portal-visibility frame).
Gate: dotnet build green; App suite 134/134; offline launch confirms both
shaders compile + link with no GL errors.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
CellView convex polygon edges → clip-space planes (nx,ny,0,d) for gl_ClipDistance,
≤8 with collinear-edge merge. Multi-polygon or >8-edge regions degrade to the union
AABB scissor (over-include, never hide); empty regions are distinct (draw nothing).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Code review caught a CRITICAL under-inclusion: ApplyReciprocalClip scanned for the
first OtherCellId match, so a cell with two portals to the same neighbour clipped both
near-side openings against the FIRST reciprocal polygon — hiding geometry through the
second opening (real on Holtburg cellar cells 0x148<->0x149). Plumb the dat's
OtherPortalId back-link through CellPortalInfo + BuildLoadedCell and index the reciprocal
directly (retail arg2->other_portal_id, 433557). Skip (degrade to over-include) when the
index is unresolvable — never clip against a guessed polygon. Adds a disjoint two-back-
portal regression test.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Clip the portal opening against the neighbour's matching back-portal polygon
before propagating, so a cell's clip region is the intersection of the opening
seen from both sides. Closes the M-4 stub in ISSUES #102. Can only tighten,
never under-include; degrades to prior behavior when no back-portal is found.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Code-review minor follow-ups: correct the CellTodoList comments (ties are LIFO,
not FIFO — an equal-distance newcomer lands at the tail and pops first, matching
retail's break-on-first-not-greater + pop-from-tail). Update ISSUES #102 to record
that U.2a closes I-1/I-2 (under-count + duplicate accumulation) via the enqueue-once
gate, narrowing the residual to diamond-topology clip-completeness (AddToCell onward
re-propagation, tracked under U.6).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
PortalVisibilityFrame gains OrderedVisibleCells (closest-first). Replace the FIFO +
MaxReprocessPerCell cap with a distance-priority queue and a grow-watermark fixpoint
(retail InsCellTodoList 433183 / AddViewToPortals 433446) so cyclic dungeon graphs
converge without duplicate-cell blow-up.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Remove IndoorCellStencilPipeline + portal_stencil shaders, RenderInsideOutAcdream,
RenderOutsideInAcdream, the A8-perf instrumentation, the cameraInsideBuilding /
ACDREAM_A8_INDOOR_BRANCH branch, and the dead EntitySet partition values. Collapse
the render branch to the default Draw(All) path (U.4a replaces it with the gated
unified pass). Keep all audited EnvCellRenderer / BuildingLoader / CellVisibility /
camera-collision fixes.
Also deleted with the partition: the two test-only walk helpers
(WbDrawDispatcher.WalkEntitiesForTest / WalkEntitiesForTestByCellIds) and their
test files (WbDrawDispatcherEntitySetTests, WbDrawDispatcherCellIdsOverloadTests),
which existed solely to exercise the removed IndoorPass/OutdoorScenery/
BuildingShells/LiveDynamic partition. EntityMatchesSet / IsShellScopedSet collapse
to the All-path constants; the set: parameter is retained as a seam for the
unified pass.
Note: the depth-clear-if-inside default-path workaround was removed per the
U.1 task list — any current indoor-wall degradation persists until a later
Phase U task lands the unified pass (expected, not a regression introduced here).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Visual verification showed the camera vibrating/bouncing when pressed against a
wall. Cause: the sweep wrote its clamped result back into _dampedEye, so the
next frame's damping lerped from the wall toward the target and the sweep
re-clamped it — a per-frame feedback loop. Retail keeps viewer_sought_position
(damped, uncollided) separate from viewer (the published collided eye). Fix:
collide into a separate publishedEye for Position/View/fade and leave _dampedEye
as the clean sought position. New regression test
Update_CollisionDoesNotCorruptDampedState (clamp-then-release → full recovery).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds a checkable "Collide Camera (spring arm)" item to the Camera submenu.
Clicking it flips CameraDiagnostics.CollideCamera, matching the live A/B
toggle pattern used for UseRetailChaseCamera. The checkmark reflects the
current flag value so state is always visible in the menu.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Both RetailChaseCamera construction sites now supply CollisionProbe with a
fresh PhysicsCameraCollisionProbe(_physicsEngine). The per-frame Update
call gains cellId: _playerController.CellId and selfEntityId:
_playerController.LocalEntityId so the probe has the correct spatial
context for sphere-sweep queries.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Add ICameraCollisionProbe? CollisionProbe { get; init; } to RetailChaseCamera.
Extend Update() with optional cellId/selfEntityId params (default 0) so all
existing callers compile unchanged. After the exponential-damping block (step 5)
and before publishing Position/View (step 6), sweep _dampedEye through the
probe when CameraDiagnostics.CollideCamera is true and a probe is wired in
(step 5b). The fade computation in step 7 then naturally uses the collided eye.
Null probe and cellId=0 both short-circuit cleanly. Three new xUnit tests
cover: probe-wired+flag-on publishes collided eye, flag-off skips probe,
null probe doesn't throw. All 30 RetailChaseCameraTests pass.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Code review found the probe passed ObjectInfoState.None; retail's
SmartBox::update_viewer calls init_object(player, 0x5c) =
IsViewer|PathClipped|FreeRotate|PerfectClip (pseudo-C :92864). PathClipped makes
the sweep hard-stop at first contact (TransitionTypes.cs:811) instead of
edge-sliding around corners (which would re-trigger the A8.F camera-cell
instability); IsViewer lets the eye pass through creatures, colliding only with
world geometry. Resolves the spec's slide-vs-stop open question. Also reset
CollideCamera in the Defaults_AreRetailValues baseline test (review: maintenance
trap). Spec §5.1/§11.1 synced.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
First-fix from the visual-gate-failure handoff: an empty OutsideView means
"no outdoors visible from here," not "all outdoors." When inside a building
with an empty clipped mask, Step 4 now draws NO terrain/scenery instead of
disabling the stencil and flooding ungated terrain over the cell interior
(the Step-3 walls already occupy the framebuffer). Visual-confirmed: Holtburg
cottage cellar walls are solid now, no terrain bleed-through.
Also adds portal diagnostics that root-caused so-called "Bug B":
- PortalVisibilityBuilder: per-camera-cell CAMPORTAL census (polyLen +
side-test result) emitted BEFORE the BFS guards, so an empty OUTSIDEVIEW
can be traced to the exact gate.
- A8CellAudit `portals`: replicate BuildLoadedCell's polygon-vertex
resolution so PortalPolygons[i] validity is checkable offline.
Finding: the builder is largely CORRECT — it produces narrowed clipped
OutsideView regions for most cells (0172/0173/0162/015E/0165/016F). The
empty cases are mostly legitimate (windowless cellar can't see out; the
3rd-person camera eye on the outdoor side of a front-door plane culls that
exit). The handoff's Finding 2 ("under-produces, never narrows") is
substantially not real. Remaining wall-missing regressions in OTHER
buildings live in the cross-building Step-5 enforcement, escalated separately.
All gated behind ACDREAM_A8_INDOOR_BRANCH=1; default play unaffected.
App tests 108/108.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Env-gated diagnostics (off by default; do not affect the default game):
- ACDREAM_A8_DUMP_PV=1: PortalVisibilityBuilder dumps local→NDC→clipped portal
geometry + OutsideView poly count for the first 2 Build calls per camera cell.
- ACDREAM_PROBE_ENVCELL=1: [opaque] line dumps the opaque cell-render stats
(cells/tris) BEFORE the per-cell transparent loop overwrites _envCellRenderer.Stats.
Used to diagnose the A8.F visual-gate failure (see handoff doc). Gated behind
ACDREAM_A8_INDOOR_BRANCH=1 like the rest of the indoor branch.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
DrawRegionBit2 set DepthFunc.Always and never restored it; EnvCellRenderer
and WbDrawDispatcher rely on ambient DepthFunc, so the leak made clipped
translucent cells, the camera cell + cells iterated after a clipped one, and
the IndoorPass building shells all render with Always instead of Less (walls
drawing through each other). DrawRegionBit2 now restores DepthFunc.Less on
exit; EnableBit2CellPass sets the per-cell render state (Less + depth-write
off) explicitly so the bug class can't silently recur. ColorMask matched to
the indoor pass (alpha-write off).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The clip predicate (w+z>=0) is convention-agnostic, not GL-specific:
Matrix4x4.CreatePerspectiveFieldOfView (which all acdream cameras use) is
NDC z in [0,1], not [-1,1]. Comment said "GL near plane / z_ndc>=-1" which
is misleading though the code is correct (eye w=0 always excluded; divide
safe under both conventions). Also soften the ProjectToNdc CCW claim: it
preserves projected winding; the caller must feed camera-facing portals.
No behavior change. (Opus code-review I-1/M-1.)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Task 0 baseline cleanup. Removes the temporary A8 step-disable diag
toggles (A8Diag* properties + ACDREAM_A8_DIAG_* env reads) that the A8
batch left behind in RuntimeOptions, and unwraps their guards in
GameWindow.RenderInsideOutAcdream so every guarded draw (Step 2 punch,
Step 3 EnvCell-opaque + IndoorPass, Step 4 terrain + outdoor scenery,
portal depth-clamp) now runs unconditionally. RuntimeOptionsTests drops
the matching assertions. The ACDREAM_PROBE_VIS apparatus
(EmitDrawOrderProbe / EmitStencilProbe / EmitBuildingsProbe /
EmitEnvCellProbe) is preserved untouched.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Lands the working A8 indoor-rendering and streaming fixes accumulated this
session. User has verified these visually to some degree (e.g. lifestone /
translucent meshes confirmed fine under the FrontFace flip; bridge / wall /
collision regressions confirmed fixed after travel); not every path has been
exhaustively gated. The cellar-flap defect remains OPEN and will be solved
the retail-faithful way via a dedicated brainstorm (see handoff docs).
Rendering core (reviewed, high confidence):
- EnvCellRenderer SSBO stride fix: upload packed Matrix4x4[] (64B) instead of
the 80B CPU InstanceData struct the shader never expected — fixes the
transform/texture "explosion" for any draw with >1 instance (cells that
dedupe to a shared cellGeomId). Real root cause.
- WB-style global FrontFace(CW) + per-batch CullMode carried through the MDI
layout (GroupKey + BuildIndirectArrays + DrawIndirectRange split into
same-cull runs with absolute uDrawIDOffset per run).
- EntitySet partitioning (IndoorPass / OutdoorScenery / LiveDynamic) +
WorldEntity.BuildingShellAnchorCellId so building shells scope to their
dat-derived building cell instead of rendering everywhere.
- RenderOutsideInAcdream (look into buildings from outside) +
CollectVisiblePortalBuildings frustum cull of portal bounds.
- Sky-when-inside-building + per-cell audit probe + GL-state probe.
Streaming / perf (test-covered; not independently code-reviewed this session):
- Near/far priority queues so near work wins over far; PromoteToNear carries
full landblock + mesh data; LandblockEntriesWithoutAnimatedIndex avoids
rebuilding the animated-lookup dict in the hot draw path. Fixes the
bridge-not-appearing / missing-walls / broken-collision-after-travel
regressions and improves post-transition FPS.
Tooling + docs:
- tools/A8CellAudit: offline dat cell/portal/building dumper (portals +
buildings modes) — reproduces the cellar-flap investigation with no launch.
- docs/research cellar-flap root-cause + option-2 handoff (the didInsideStencil
double-duty finding + the WB-recursive design decision + brainstorm prompt),
entity-taxonomy, replan, issue-78 visibility investigation.
Diagnostics retained on purpose: ACDREAM_A8_DIAG_* gates, portal_stencil.vert
provisional pos.w clamp, and the probe families are kept (env-var gated, zero
cost when off) because the pending option-2 cellar-flap brainstorm needs them.
Strip in the option-2 ship commit.
Indoor branch stays behind ACDREAM_A8_INDOOR_BRANCH=1 (default off = pre-A8
visual). Build green; App tests + Core (streaming/dispatcher/loader) tests pass.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Visual-gate-#4 evidence revealed the prior commit's cull-restore-at-exit
addition was wrong. The Landblock→None CullMode override worked correctly
for cell-mesh polys, but the cull-back state I restored at Render exit
propagated to the subsequent `dispatcher.Draw(IndoorPass)` call. The
dispatcher's IndoorPass renders AC's cottage shell — landblock-baked
GfxObj parts (wooden floor planks, wall slabs) whose pos-side winding +
our FrontFace=CCW + cull-back = floor poly is back-facing and culled.
User saw light blue sky through the floor in gate-#4.
Reverting the cull-restore lets cull-disabled propagate from
EnvCellRenderer.Render through IndoorPass. Cottage shell renders
double-sided so the floor + wall slabs are visible from any angle.
Step 4's gl.Enable(EnableCap.CullFace) at the terrain pass (line
~10768) + the cleanup block's enable (line ~10870) re-establish
cull-back BEFORE the LiveDynamic dispatcher.Draw fires — so chars,
NPCs, doors still render solid (no see-through-head regression from
gate-#3's ACDREAM_A8_DISABLE_CULL=1 diagnostic).
The retail-faithful long-term fix is matching WB's `glFrontFace(GLEnum.CW)`
globally (per GameScene.cs:843) so cull-back selects the correct side
for AC's natural polygon winding without needing double-sided rendering.
That requires a wider audit of every consumer's FrontFace assumption
(translucent crystal renderer + others) and is deferred.
14/14 EnvCellRenderer tests pass. Build green.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The cull A/B diagnostic (prior commit's ACDREAM_A8_DISABLE_CULL=1) in
visual-gate-#3 confirmed: cell-mesh polys are being culled by back-face
culling, which is why floors disappear when looking down from inside a
room. Per-cell audit data showed every cell-mesh batch has
CullMode.Landblock — assigned because AC's CellStruct polys carry
SidesType=Landblock in the dat. Our SetCullMode maps Landblock to
glCullFace(Back), matching WB.
Root cause:
WB sets `glFrontFace(GLEnum.CW)` globally at GameScene.cs:843. Our
WbDrawDispatcher.cs:1056 sets `glFrontFace(CCW)` — the GL default,
opposite of WB. With our flipped-from-natural fan triangulation in
BuildCellStructPolygonIndices (which emits (i, i-1, 0) for each fan
triangle, reversing the input vertex order), the resulting effective
winding from the camera's perspective is OPPOSITE WB's. Cull-back then
removes the OPPOSITE face from what WB does — hiding the floor side
that should be visible from inside the room.
Within a single cell-mesh batch, the polys face every direction (walls
outward, floor up, ceiling down) but all share CullMode.Landblock. No
single cull setting can be correct for all three orientations
simultaneously — the retail-faithful approach is to render cell polys
double-sided (cull off).
Two changes scoped to EnvCellRenderer.RenderModernMDIInternal so other
renderers aren't affected:
1. Remap CullMode.Landblock → None when iterating per-cull-mode
batch groups. Cell polys render with cull disabled, all faces
visible. CullMode.Landblock is only assigned by
PrepareCellStructMeshData (cell polys) in this codebase — terrain
uses a different render path. Scope is exactly right.
2. Explicitly Enable(CullFace) + CullFace(Back) at Render exit so the
dispatcher's subsequent IndoorPass + LiveDynamic Draws don't
inherit the cull-disabled state. The see-through-head symptom in
visual-gate-#3 was caused by exactly this state leak from the
ACDREAM_A8_DISABLE_CULL=1 diagnostic; the proper fix needs the
explicit restore. Also updates the static `_currentCullMode` cache
so the next Render call's first SetCullMode comparison is correct.
Removed the ACDREAM_A8_DISABLE_CULL diagnostic env var — its role as
A/B test is complete. 14/14 EnvCellRenderer tests pass. Build green.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two changes from visual-gate-#2 evidence.
LiveDynamic fix (real bug closure):
The user reported "can't see char ... door is missing" in visual gate #2.
Doors and the player char are LiveDynamic entities (ServerGuid != 0). The
outdoor branch's Draw(set: All) includes them; the indoor branch's
RenderInsideOutAcdream only renders IndoorPass + OutdoorScenery partitions,
implicitly excluding LiveDynamic. The method's own header comment promised
"LiveDynamic is drawn last in BOTH branches" but no call existed in the
indoor path — a documented behavior with no implementation. Wire the
LiveDynamic Draw after RenderInsideOutAcdream returns with stencil + state
restored to defaults at its cleanup block.
Cull A/B diagnostic (bisect floor-missing root cause):
ACDREAM_A8_DISABLE_CULL=1 forces every cell-mesh batch's effective CullMode
to None. The visual-gate-#2 audit confirmed cell meshes upload correctly
(every cell has multi-batch render data with non-zero indices, no null
data, no zero handles). Every batch uniformly reports CullMode.Landblock
which maps to glCullFace(Back) — identical to WB's mapping. So data is
fine and CullMode lookup is fine; only the BIND-TIME interaction (polygon
winding orientation in our coord system + cull-back) could still hide
specific polys. If floor appears with this gate set, cull/winding is the
remaining bug (need to either invert winding upstream or remap CullMode);
if not, the issue is elsewhere (lighting / depth / alpha) and we look
there. Tight bisect — one launch's evidence.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
