# Task C - ribs / structural intrusions, C02 shell vs INT R02 (PI-BLD-0001-V01)

Read-only audit. Nothing under C:\q2\r or the project was modified. Labels: measured / proposed / interpreted.

## Method (measured unless stated)
- GLB loaded with a copy of the refit.py loader (glb_load.py); glTF (x,y,z) -> source (x,-z,y), front = -Y.
- Parts sliced by horizontal planes (exact triangle/plane intersection, endpoints rounded to 1e-4 m). Closed manifold parts (Rib_*, Gold_inner_*, Mullion_*, atrium golds, cap) are filled and intersected with the room rectangle (area, footprint). Open-surface parts (Tower_curtainwall_skin, Podium_wall_*, Storey_band_*, Front_amber_atrium_skin, Cantilever_fascia_*) have no volume; only their slice trace (length) is tested.
- Heights per room: z_bottom+0.1, mid, z_top-0.1 (as asked) plus a 0.25 m sweep from z_bottom+0.1 to z_top-0.1. z_bottom = z_m, z_top = z_m + clear_height_m.
- Families: RIB = Rib_R*, Gold_inner_R*, Crown_glass_R*; MULL = Mullion_*, Storey_band_*; ATRIUM = Front_amber_atrium_skin, Rear_crossed_gold*, Front_atrium_gold*, Central_glazed_cap; ENCL (context only) = tower skin, podium walls, arches, fascia, canopy, terraces, basins. Slabs (Foundation, Deck_*, Occupied_roof) excluded.
- "Depth" = max distance from the room boundary reached by the footprint inside the room (footprint vertices densified to 0.05 m).
- Door opening = segment of length clear_width_m centred at (x,y) along `axis`; door height range = level z .. z + clear_height_m (0.25 m sweep). "Open width after" = width minus the larger of (segment covered length, projection of obstacles inside a 1.0 m approach box on each side).
- Capsule sweep (0.6/0.7/0.8/1.0 m diameters): room rectangle eroded by d/2 minus obstacle buffered by d/2, over the full room height envelope; plus plan gap from obstacle envelope to each room edge. The real capsule diameter and height are UNKNOWN; this is a sweep only.
- Splitting: rectangle minus obstacle (and capsule-eroded variant), connected components with area > 0.001 m2, checked at every sampled height.

## Results
1. **Intrusions (measured).** Only L03-RM001 is intruded by a rib/mullion/band/atrium part: Rib_R1. Enters at z=22.00 (4.00 m above the floor at z=18), up to the ceiling z=23.40. Slice area in room: 0.085 m2 @ z22.25, 0.551 @22.75, 1.192 @23.25, 1.403 @23.39 (max). Plan envelope 1.417 m2, x -38.00..-37.12, y -40.42..-37.89. Max depth 0.873 m at the ceiling plane (0.816 m at z=23.3). Overhead only (4.0 to 5.4 m above floor). Nearest door L03-D001: 22.195 m; no door within 1.0 m. Tower rooms L25/L26 have no intruding part.
2. **Vs published `rib_intrusions` (measured).** Published = {L03-RM001: {Rib_R1: 26}}, all else {}. Agreement 81/81 rooms, 0 disagreements. 26 is a surface-sample count, not an area. The published test only looks at `Rib_*`; I also tested Gold_inner, Mullion, Storey_band and atrium parts, no extra room. With only the three requested heights, L03-RM001 is seen only at z_top-0.1 (23.3).
3. **Doors (measured).** 102 doors: no rib/mullion/band blocks or narrows any opening; none within 1.0 m. Closest rib-family part to any door 7.884 m (L26-D011, Gold_inner_R4), next 9.903 m (L25-D006, Rib_R6). Closest mullion/band 4.512 m (L26-D010, L26-D011, Storey_band_18). Closest atrium part 5.364 m (L26-D004). Per-door numbers: doors_102.csv. Side finding (not a rib): L00-D001 (portal, 8 m wide) has Podium_wall_1 (y=-52.4, no cut-out) 0.4 m outside it across the full width (the tool flags BLOCKS in the ENCL column).
4. **L03-RM001 vs Rib_R1 shortfall (measured).** Room x -38..-14, y -48..-24. Room-side shortfall: the rib reaches 0.873 m into the room at the ceiling. Smallest rigid shift that clears Rib_R1 over the room's z range: +x 0.880 m; +y (rearward) 10.115 m; -y 16.425 m; -x 30.005 m. The README's ~4.8 m rearward relocation (for the wall) would still overlap Rib_R1 by 1.417 m2 (measured by shifting the rectangle). L03-RM001 also lies across Podium_wall_2 (y=-44.4, 3.6 m inside the room edge) - context.
   **L25/L26 capsule sweep (measured, interpreted as clear width):** ribs are entirely outside all 32 tower rooms; closest rib-family part 2.256 m (L26-RM001 Rib_R5), 2.257 (L26-RM002 Rib_R6), 2.405 (L26-RM003 Rib_R1, L26-RM006 Rib_R2). Closest of any rib/mullion/band/atrium part: 1.456 m (L26-RM005, Front_atrium_gold_-1), 1.457 (L26-RM004, Front_atrium_gold_1), 1.505 (L26-RM005, Mullion_51), 1.613 (L26-RM004, Storey_band_18). No L25/L26 room has a gap under 0.6, 0.7, 0.8 or 1.0 m. Outside L25/L26: L00-RM001 is 0.6 m from Front_amber_atrium_skin (below the 0.8 and 1.0 m sweep values).
5. **Splitting (measured).** No part splits any room into disconnected pieces (raw or capsule-eroded). L03-RM001 with the Rib_R1 notch removed: 1 piece, 574.58 of 576 m2.

## Not tested / unknown
- Real capsule diameter, height, and whether the player walks outside the room rectangle: unknown. Sweep only.
- The rib blades are shell geometry with no collision, thickness assembly or structural data; "intrusion" is geometric overlap only, not a structural check.
- Heights between sampled slices are covered at 0.25 m spacing (Rib_R1 entry height refined to 0.01 m); features thinner than 0.25 m in z could be missed (none expected for ribs).
- Open surfaces (skin, walls, storey bands, fascia, amber skin) have no thickness in the mesh; their intrusion is a line, not an area.
- Door swing arcs, door leaf positions, furniture, and rooms on levels without layouts were not tested. Rib blades were not checked against the building's own enclosure beyond the room rectangles.
- The INT R02 `dimension_status` is PROPOSED; room rectangles are planning zones, wall thickness indicative.

## Files
findings.csv / findings.json (10 findings), rooms_81.csv, doors_102.csv, task1_room_intrusions.json, task1b_nearest_by_room.json, task3_doors.json, task4a_L03RM001_RibR1.json, task4b_capsule_sweep_all_rooms.json, task5_room_splits.json. Code: glb_load.py, analysis.py + analysis2.py (concatenated in run_all.py), make_outputs.py.
