 |
I made a partial fix in svn rev 8950. I simply masked the load disabling for CPU-resident images: void idImage::ActuallyLoadImage( bool allowBackground ) { ... if ( session->IsFrontend() && !(residency & IR_CPU) ) { common->Printf( "Trying to load image %s from frontend, deferring...\n", imgName.c_str() ); Unfortunately, it is logically wrong for IR_BOTH images, since they can now get loaded to GL from frontend thread. But I'm not sure we have any such images. Default image is the only IR_BOTH, and it gets loaded from backend. To be honest, I did not remember ever seeing this message in game console. |
|
|
Also committed a bit of refactoring in svn rev 8949. Before that cubemaps were loaded straight in idImage::ActuallyLoadImage. Now they are loaded in two steps just like ordinary textures: R_LoadImageData: loads from file on disk to cpuData R_UploadImageData: loads from cpuData to GL and purges cpuData (customizable via residency flag) |
|
|
While I did not do what the title of the issue says, by original problem with cutoffTimeMap and collisionStatic maps is fully fixed. So I'll mark this issue as resolved. |
- Anonymous
