The current implementation of collision detection does not actually handle bounding boxes, only a dummy implementation is provided, like:
return { -10000, 10000, -10000, 10000 };
Which is problematic, because the method returns the result by const reference:
virtual const geo::EmptyFrame& GetBoundingBox() = 0;
But it's never passed to a const reference variable to extend its lifetime. This happens to not cause any problems in debug builds, but it does with -O2 optimization level, and results in polygon collisions being undetectable.
The current implementation of collision detection does not actually handle bounding boxes, only a dummy implementation is provided, like:
return { -10000, 10000, -10000, 10000 };Which is problematic, because the method returns the result by const reference:virtual const geo::EmptyFrame& GetBoundingBox() = 0;But it's never passed to a const reference variable to extend its lifetime. This happens to not cause any problems in debug builds, but it does with-O2optimization level, and results in polygon collisions being undetectable.