Parse bounding vertices, triangles, and normals from M2 files and use
them for proper triangle-level collision instead of AABB heuristics.
Spatial grid bucketing for efficient queries, closest-point wall push
with soft clamping, and ray-triangle floor detection alongside existing
AABB fallback.
- Add mutex to AssetManager::loadTexture/loadDBC/fileExists to prevent
StormLib thread-safety races that silently fail texture reads; stop
caching texture load failures so transient errors are retried.
- Replace /unstuckgy DBC lookup (which used wrong coordinate transform)
with hardcoded safe locations per map.
- Widen WMO floor raycast from single grid cell to ±1 unit range query
to catch bridge/walkway triangles at cell boundaries.
- Tighten swept collision hit threshold (0.5 → 0.15) and grid query
margin (2.5 → 1.5) to prevent false-positive wall pushes.
- Tighten post-wall-push Z snap lower bound (-1.0 → -0.3) to prevent
gradual floor sinking.
The particle emitter parser had three bugs causing OOM crashes during loading:
- Struct size was 496 bytes instead of correct WotLK 476 (0x1DC), misaligning multi-emitter models
- FBlocks were read as 20-byte M2TrackDisk instead of 16-byte FBlockDisk (no interp/seq prefix)
- parseAnimTrack had no cap on sequence count, allowing garbage data to allocate billions of entries
Replace flat mesh rendering of additive/mod blend batches (blendMode >= 3)
with camera-facing point sprites using a soft radial gradient texture and
additive blending. Adds M2 particle emitter infrastructure (structs, shader,
parsing stubs) but disables emitter parsing — the assumed 476-byte struct
size is wrong for WotLK 3.3.5a, causing misaligned reads that explode RAM.
Water/lava batches in fountain and Ironforge M2 models use non-opaque
blend modes (alpha, additive) defined in the M2 material table. Without
parsing these, they rendered as solid surfaces extending visibly beyond
their containers. Now each batch looks up its blend mode from the
material array and sets the appropriate GL blend function.
Parse M2TextureTransform entries and texture transform lookups from the
M2 binary, then apply per-batch UV offsets in the vertex shader using
the existing animation time base and global sequence durations.
M2 batches have a geosetIndex field that determines the geoset group for
filtering (0-18=body, 1xx=hair, etc.). We were incorrectly using submeshId
from the submesh struct instead. Now properly stores and uses geosetId.
- Use explicit loaded flag for heightmap instead of checking height values
- Increase player collision radius from 0.50 to 0.55 for better wall collision
Portal-based visibility culling for WMO rendering (disabled by default,
needs debugging for complex WMOs like Stormwind). Skip character creation
screen when characters already exist in single-player mode.
- Parse global sequence durations from M2 binary and use them in bone
interpolation so torches, candles, and other env doodads animate.
- Add UV scroll shader effect for smoke models (HouseSmoke, SmokeStack)
as a workaround for unimplemented M2 particle emitters.
- Tighten WMO floor probe heights to prevent multi-story buildings from
returning the wrong floor, fixing player clipping through inn floors
and camera locking onto the second floor.
- Use player ground level as reference for camera orbit floor collision
so the camera doesn't fight upper floors in buildings.
The MOHD header structure was missing the nTextures field at the start,
causing all subsequent field reads to be offset by 4 bytes. This corrupted
bounding box values and caused floor geometry to be incorrectly culled.
