Call of Duty Brushes

[snake-biscuits]

Was looking at kungfooman/CoD-Bsp-Decompiler recently
Thought there might be a few structs in there I don't have, and there were.

Brushes are extremely barebones, not unlike respawn.titanfall

struct Brush {
  uint16_t num_sides;
  uint16_t unknown;
};

struct BrushSide {
  union { float distance; uint32_t index; } plane;
  uint32_t texture;
};

[snake-biscuits]

To index BrushSides you need to keep track of those indexed by previous Brushes
In respawn.titanfall CMBrush has some members to keep track of this

Axial sides do not have planes (just like respawn.titanfall)
The plane union switches between float plane distance & uint32_t plane index for axial & non-axial sides.

Comparing the mp/mp_lobby.d3dbsp test map to mp_lobby.map seems to indicate the following axial order:

• -X +X -Y +Y -Z +Z

Haven't 100% confirmed this, but it seems to line up
Brush & side order isn't the same between the .d3dbsp & .map

[snake-biscuits]

.d3dbsp

>>> import bsp_tool
>>> bsp = bsp_tool.load_bsp("tests/maps/Call of Duty 4/mp/mp_lobby.d3dbsp")
>>> bsp.brush(0).sides
[BrushSide(Plane(vec3(-1.0, 0.0, 0.0), 192.0), 'caulk', ...),
 BrushSide(Plane(vec3(1.0, 0.0, 0.0), 192.0), 'caulk', ...),
 BrushSide(Plane(vec3(0.0, -1.0, 0.0), -176.0), 'case512grey', ...),  # -Y
 BrushSide(Plane(vec3(0.0, 1.0, 0.0), 240.0), 'caulk', ...),  # +Y
 BrushSide(Plane(vec3(0.0, 0.0, -1.0), 192.0), 'caulk', ...),
 BrushSide(Plane(vec3(0.0, 0.0, 1.0), 192.0), 'caulk', ...)]

.map

>>> from bsp_tool.extensions.editor.map.cod4 import MapFile as CoD4MapFile
>>> map_ = CoD4MapFile.from_file("tests/mapsrc/CoD4Radiant/Call of Duty 4/mp_lobby.map")
>>> [i for i, b in enumerate(map_.entities[0].brushes) if 240 in [abs(bs.plane.distance) for bs in b.sides]]
[0, 5]
>>> map_.entities[0].brushes[5].sides
[BrushSide(Plane(vec3(-0.0, -0.0, -1.0), 192.0), 'caulk', ...),
 BrushSide(Plane(vec3(-0.0, -0.0, 1.0), 192.0), 'caulk', ...),
 BrushSide(Plane(vec3(-0.0, -1.0, -0.0), -176.0), 'case512grey', ...),  # -Y
 BrushSide(Plane(vec3(1.0, -0.0, -0.0), 192.0), 'caulk', ...),
 BrushSide(Plane(vec3(-0.0, 1.0, -0.0), 240.0), 'caulk', ...),  # +Y
 BrushSide(Plane(vec3(-1.0, -0.0, -0.0), 192.0), 'caulk', ...)]

[snake-biscuits]

Oddly, the plane distance is inverted for planes on negative axes
call_of_duty.modern_warfare & respawn.titanfall likely have a different approach to plane sided-ness
Both abandon the quake "store each plane & it's inverse in pairs" system

Presumably to save on map size
Or taking advantage of modern hardware & abandonding the optimisation

[snake-biscuits]

Oddly, the plane distance is inverted for planes on negative axes

Incorrect, maybe I messed something up with loading planes from the .map format
This should hold true:

mins = sum((s.plane.normal * s.plane.distance for s in brush.sides if -1 in s.plane.normal), start=vec3())
maxs = sum((s.plane.normal * s.plane.distance for s in brush.sides if +1 in s.plane.normal), start=vec3())

btw you should be able to find points on planes with this approach
sum of any 3 normal * distance; dot product between each pair of normals must be > 0

[snake-biscuits]

This brush system might also be used in older infinity_ward branches
Can check that when I have the new modern_warfare.brush method on my PC w/ all the maps.

[snake-biscuits]

Accidentally linked the commit to #176
oops

[snake-biscuits]

CallofDuty2ModTools.exe install includes test maps (map_source/*.map, devraw/maps/*.d3dbsp)
This means I can compare raw .map to .d3dbsp to confirm findings
Assuming the .map format isn't too different...