mirror of
https://github.com/Kelsidavis/WoWee.git
synced 2026-05-06 09:03:52 +00:00
a845723635
ROOT CAUSE: createInstance() checks mdlRef.isValid() which requires vertexBuffer != VK_NULL_HANDLE. But vertex buffers are uploaded via staging and only finalized by waitAllUploads(). Instances were being created BEFORE the upload flush, so vertexBuffer was still null, cachedIsValid was set to false, and all instances were skipped during render (0 draws despite loaded models). Fix: split rebuildObjects into two phases: 1. Load all models (upload geometry to staging) 2. waitAllUploads + pollUploadBatches (finalize GPU buffers) 3. Create all instances (vertexBuffer is now valid, isValid() = true) This matches the client's terrain_manager pattern where models are loaded on background threads and instances created after finalization. |
||
|---|---|---|
| .. | ||
| asset_extract | ||
| auth_login_probe | ||
| auth_probe | ||
| blp_convert | ||
| dbc_to_csv | ||
| editor | ||
| asset_pipeline_gui.py | ||
| backup_assets.sh | ||
| diff_classic_turtle_opcodes.py | ||
| gen_opcode_registry.py | ||
| generate_ffx_sdk_vk_permutations.sh | ||
| m2_viewer.py | ||
| opcode_map_utils.py | ||
| validate_opcode_maps.py | ||