Merge pull request 'test(M00): diagnose checkpoint camera occluders' (#5) from work/sindo-main-codex/m00-camera-occluders into master

Reviewed-on: #5

coordination/claims/M00-QAR-CAMERA-OCCLUDERS-001.md

@@ -0,0 +1,78 @@
+# M00-QAR-CAMERA-OCCLUDERS-001 — Camera occluder diagnostic
+
+<!-- OPENWC_CLAIM:M00-QAR-CAMERA-OCCLUDERS-001:sindo-main-codex:2026-07-13 -->
+
+## Ownership
+
+- Target: M00
+- Program: QAR
+- Owner/Agent ID: sindo-main-codex
+- Branch: `work/sindo-main-codex/m00-camera-occluders`
+- Lease expires UTC: 2026-07-13
+- Integrator: milestone integrator
+
+## Outcome
+
+Report scene-tree M2/WMO geometry that contains calibrated cameras or intersects each camera-to-target segment.
+
+## Non-goals
+
+- Changing placements, cameras, FOV, culling, or renderer behavior.
+- Claiming full coverage of RenderingServer RID-only instances.
+- Implementing a production visibility or collision service.
+
+## Paths
+
+- Exclusive: `src/tools/probe_render_camera_occluders.gd`
+- Shared/hotspots: renderer baseline documentation
+- Generated/ignored: local JSON probe reports
+
+## Contracts and data
+
+- Public API/events: headless diagnostic CLI only
+- Schema/format version: report schema 1
+- Migration/compatibility: none
+- Consumers: M00 fidelity workflow
+
+## Dependencies
+
+- Requires: calibrated five-point manifest and streaming scene
+- Blocks: placement versus camera-composition classification
+- External state: local extracted/cache data
+
+## Verification
+
+- Commands: camera occluder probe, coordination and documentation gates
+- Fixtures: five calibrated camera/target segments
+- Fidelity evidence: scene-tree bounds at paired build 12340 viewpoints
+- Performance budget: offline diagnostic
+
+## Documentation deliverables
+
+- Inline public API docs: probe header and output fields
+- Module specification: verification/source map
+- Data-flow diagram: occluder probe flow
+- Sequence/state/dependency diagrams: synchronous diagnostic; not applicable
+- Source map/status updates: baseline findings
+
+## Simplicity and naming
+
+- Important names introduced: `camera_containing_geometry`, `segment_intersecting_geometry`
+- Simplest considered solution: transformed published AABBs
+- Rejected complexity/abstractions: GPU visibility readback or new BVH
+- Unavoidable complexity and justification: RID-only geometry cannot be named by this probe
+- Measured optimization evidence: not applicable
+
+## Status
+
+- State: ready
+- Done: five-point scene-tree AABB probe, containment/intersection classification and documentation
+- Next: integrator review; calibrate reproducible reference camera direction/FOV separately
+- Blocked by:
+
+## Handoff
+
+- Commit: branch HEAD
+- Results: zero containing geometry at all five cameras; expected WMO/liquid target intersections; ADT/dense segments unobstructed
+- Remaining risks: RID-only instances are excluded; manual reference direction and FOV were not recorded exactly
+- Documentation updated: `docs/RENDER_BASELINE.md`, `docs/modules/world-renderer.md`

docs/RENDER_BASELINE.md

@@ -158,3 +158,24 @@ flowchart LR
 Измеренный clearance: terrain overview `89.044`, ADT boundary `44.788`, dense M2 `90.178`, large WMO `12.034`, waterfall примерно `76.128` Godot units. Следовательно, все пять камер находятся над rendered terrain; visual obstruction принадлежит placements/WMO/composition, а не terrain height.
 
 Waterfall XZ сначала давал `no_intersection`, хотя tile `30_49` был available, полностью загружен, имел `control_splat_cache` quality mesh и LOD0 mesh, а probe находился внутри mesh AABB. Ray со смещением `2.0` units пересёк тот же mesh на высоте `113.872`; точная XZ попала на triangle seam/edge numerical miss. Probe теперь сообщает `sampled_nearby`, distance и source tile вместо ложного streaming ownership gap. `--require-all` остаётся строгим режимом для действительно неснятых точек.
+
+## Camera occluder diagnostic
+
+Scene-tree placement composition проверяется transformed AABB без изменения renderer:
+
+```powershell
+godot --headless --path . --script res://src/tools/probe_render_camera_occluders.gd -- --wait 3
+```
+
+```mermaid
+flowchart LR
+    C[Calibrated camera] --> A[Published Mesh/MultiMesh AABBs]
+    T[Manifest target] --> S[Camera-to-target segment]
+    A --> P[Camera containment test]
+    A --> I[Segment intersection test]
+    S --> I
+    P --> J[JSON occluder report]
+    I --> J
+```
+
+Ни одна из пяти камер не находится внутри опубликованной scene-tree geometry. Terrain-overview segment пересекает четыре Stormwind WMO groups, large-WMO segment — три Goldshire Inn groups, waterfall segment — liquid surface; ADT boundary и dense-M2 segments не пересекают placement AABB. Поэтому прежнее визуальное впечатление «камера внутри WMO/placements» не подтверждается. Основной paired gap сейчас — неточно воспроизведённые manual look direction/target/FOV reference-кадров: например, автоматический Goldshire target направляет луч через фасад внутрь WMO, тогда как reference был вручную скадрирован на весь фасад. Probe охватывает только scene-tree MeshInstance/MultiMesh; RID-only instances не имеют доступного semantic path и явно исключены из coverage.

docs/modules/world-renderer.md

@@ -224,6 +224,7 @@ Exact exported settings and cache versions remain documented in [`../../RENDER.m
 - Original-client paired fidelity evidence incomplete.
 - Первый paired run выявил coordinate/placement mismatch: несколько совпадающих server-derived camera positions оказываются под terrain или внутри WMO/rocks OpenWC.
 - Terrain-height probe исключил under-terrain состояние для всех пяти точек; waterfall exact-XZ miss классифицирован как TriangleMesh seam/edge и подтверждён nearby sample в 2 units.
+- Camera-occluder probe не нашёл camera containment в пяти точках; paired mismatch локализован прежде всего в manual look direction/target/FOV calibration, с явным ограничением по RID-only geometry.
 - D3D12 descriptor and shutdown RID/resource issues remain.
 - M2/WMO/material/particle/ribbon/portal parity incomplete.
 - Public API is mostly exported configuration rather than stable contracts.
@@ -245,6 +246,7 @@ Exact exported settings and cache versions remain documented in [`../../RENDER.m
 | `src/tools/capture_render_checkpoints.gd` | Deterministic no-roll checkpoint camera, performance and visual capture |
 | `src/tools/verify_render_coordinate_calibration.gd` | Build 12340 camera-coordinate golden point round-trip diagnostic |
 | `src/tools/probe_render_terrain_height.gd` | Offline active-mesh terrain height and camera-clearance report |
+| `src/tools/probe_render_camera_occluders.gd` | Scene-tree placement containment and camera-to-target AABB intersection report |
 
 ## Related decisions and references
 

src/tools/probe_render_camera_occluders.gd

@@ -0,0 +1,166 @@
+extends SceneTree
+
+## Reports published scene-tree geometry around calibrated renderer cameras.
+## Usage: godot --headless --path . --script res://src/tools/probe_render_camera_occluders.gd --
+##   [--wait 3.0] [--output user://render_camera_occluders/report.json]
+
+const MANIFEST_PATH := "res://src/tools/render_baseline_manifest.json"
+const MAX_REPORTED_INTERSECTIONS := 20
+
+
+func _initialize() -> void:
+	_run_async.call_deferred()
+
+
+func _run_async() -> void:
+	var arguments := OS.get_cmdline_user_args()
+	var wait_seconds := float(_argument(arguments, "--wait", "3.0"))
+	var output_path := _argument(arguments, "--output", "user://render_camera_occluders/report.json")
+	var only_filter := _argument(arguments, "--only", "").to_lower()
+	var manifest := _load_json(MANIFEST_PATH)
+	var packed_scene: PackedScene = load(String(manifest.get("scene", "")))
+	if packed_scene == null:
+		push_error("CAMERA_OCCLUDER_PROBE: cannot load streaming scene")
+		quit(1)
+		return
+	var world := packed_scene.instantiate() as Node3D
+	var camera := Camera3D.new()
+	camera.name = "OccluderProbeCamera"
+	camera.current = true
+	world.add_child(camera)
+	world.set("camera_path", NodePath(camera.name))
+	world.set("debug_streaming", false)
+	get_root().add_child(world)
+	await process_frame
+	await process_frame
+
+	var results: Array[Dictionary] = []
+	for checkpoint_variant in manifest.get("checkpoints", []):
+		if not (checkpoint_variant is Dictionary):
+			continue
+		var checkpoint: Dictionary = checkpoint_variant
+		if not checkpoint.has("reference_wow_camera"):
+			continue
+		var checkpoint_name := String(checkpoint.get("name", "checkpoint"))
+		if not only_filter.is_empty() and not checkpoint_name.to_lower().contains(only_filter):
+			continue
+		var camera_position := _vector3(checkpoint.get("camera", []))
+		var target_position := _vector3(checkpoint.get("target", []))
+		camera.global_position = camera_position
+		if world.has_method("_refresh_streaming_targets_at"):
+			world.call("_refresh_streaming_targets_at", camera_position, true)
+		await create_timer(maxf(0.1, wait_seconds)).timeout
+		var geometry_nodes: Array[Node3D] = []
+		_collect_geometry_nodes(world, geometry_nodes)
+		var containing: Array[Dictionary] = []
+		var intersecting: Array[Dictionary] = []
+		for geometry_node in geometry_nodes:
+			var world_aabb := _geometry_world_aabb(geometry_node)
+			if world_aabb.size.is_zero_approx():
+				continue
+			var record := _geometry_record(geometry_node, world_aabb, camera_position)
+			if world_aabb.has_point(camera_position):
+				containing.append(record)
+			var intersection = world_aabb.intersects_segment(camera_position, target_position)
+			if intersection != null:
+				record["intersection_distance"] = camera_position.distance_to(intersection as Vector3)
+				intersecting.append(record)
+		intersecting.sort_custom(func(a: Dictionary, b: Dictionary) -> bool:
+			return float(a.intersection_distance) < float(b.intersection_distance))
+		if intersecting.size() > MAX_REPORTED_INTERSECTIONS:
+			intersecting.resize(MAX_REPORTED_INTERSECTIONS)
+		var result := {
+			"name": checkpoint_name,
+			"geometry_node_count": geometry_nodes.size(),
+			"camera_containing_geometry": containing,
+			"segment_intersecting_geometry": intersecting,
+		}
+		results.append(result)
+		print("CAMERA_OCCLUDERS name=%s geometry=%d containing=%d segment=%d" % [
+			checkpoint_name, geometry_nodes.size(), containing.size(), intersecting.size()])
+
+	var report := {
+		"schema_version": 1,
+		"created_utc": Time.get_datetime_string_from_system(true, true),
+		"wait_seconds": wait_seconds,
+		"coverage": "scene_tree_only; RenderingServer RID-only instances are excluded",
+		"results": results,
+	}
+	if not _write_json(output_path, report):
+		quit(1)
+		return
+	world.queue_free()
+	quit(0 if not results.is_empty() else 1)
+
+
+func _collect_geometry_nodes(node: Node, output: Array[Node3D]) -> void:
+	if node is MeshInstance3D:
+		var mesh_instance := node as MeshInstance3D
+		if mesh_instance.mesh != null and not mesh_instance.name.begins_with("TileLOD"):
+			output.append(mesh_instance)
+	elif node is MultiMeshInstance3D:
+		var multimesh_instance := node as MultiMeshInstance3D
+		if multimesh_instance.multimesh != null:
+			output.append(multimesh_instance)
+	for child in node.get_children():
+		_collect_geometry_nodes(child, output)
+
+
+func _geometry_world_aabb(node: Node3D) -> AABB:
+	if node is MeshInstance3D:
+		return node.global_transform * (node as MeshInstance3D).mesh.get_aabb()
+	if node is MultiMeshInstance3D:
+		return node.global_transform * (node as MultiMeshInstance3D).multimesh.get_aabb()
+	return AABB()
+
+
+func _geometry_record(node: Node3D, world_aabb: AABB, camera_position: Vector3) -> Dictionary:
+	var node_path := String(node.get_path())
+	var category := "geometry"
+	if node_path.contains("/M2s/") or node is MultiMeshInstance3D:
+		category = "m2"
+	elif node_path.contains("/WMOs/"):
+		category = "wmo"
+	return {
+		"category": category,
+		"node_path": node_path,
+		"distance_to_center": camera_position.distance_to(world_aabb.get_center()),
+		"aabb_position": _vector3_array(world_aabb.position),
+		"aabb_size": _vector3_array(world_aabb.size),
+	}
+
+
+func _vector3(value_variant) -> Vector3:
+	if not (value_variant is Array) or value_variant.size() != 3:
+		return Vector3.ZERO
+	return Vector3(float(value_variant[0]), float(value_variant[1]), float(value_variant[2]))
+
+
+func _vector3_array(value: Vector3) -> Array[float]:
+	return [value.x, value.y, value.z]
+
+
+func _argument(arguments: PackedStringArray, name: String, default_value: String) -> String:
+	var index := arguments.find(name)
+	if index >= 0 and index + 1 < arguments.size():
+		return arguments[index + 1]
+	return default_value
+
+
+func _load_json(path: String) -> Dictionary:
+	var file := FileAccess.open(path, FileAccess.READ)
+	if file == null:
+		return {}
+	var parsed = JSON.parse_string(file.get_as_text())
+	return parsed if parsed is Dictionary else {}
+
+
+func _write_json(path: String, value: Dictionary) -> bool:
+	var absolute_path := ProjectSettings.globalize_path(path)
+	if DirAccess.make_dir_recursive_absolute(absolute_path.get_base_dir()) != OK:
+		return false
+	var file := FileAccess.open(absolute_path, FileAccess.WRITE)
+	if file == null:
+		return false
+	file.store_string(JSON.stringify(value, "  "))
+	return true