@iche033, @scpeters, @ahcorde. Please, let me know if extra changes are required. If not, I will create a follow-up PR to add link AABB computation to the physics system update callback.

@iche033, @scpeters, @ahcorde. Please, let me know if extra changes are required. If not, I will create a follow-up PR to add link AABB computation to the physics system update callback.

took a quick look at the logic for enabling and computing AxisAlignedBox in the Link class. I think overall the approach looks fine. I see that the AxisAlignedBox component is first initialized with results computed from SDF, then later the component will be updated with the value from the physics system. I'm curious to see how similar or different the two values are. If they are very different, it could be confusing to the users. We should have a test to verify that the values are similar within some reasonable tolerance.

I'm curious to see how similar or different the two values are. If they are very different, it could be confusing to the users.

I chose to always initialize it using the SDF computation as a way to minimize inconsistencies, particularly before the physics system's first update call. This is especially relevant during the configuration of various plugins loaded from the SDF. For example, initializing it with math::AxisAlignedBox() would be ineffective for an external system performing AABB checks during its configuration step, as it would have to wait for the physics system to step at least once before obtaining a meaningful value (making plugins somewhat "dependents" of each other).

That said, if preferred, I can modify the initialization to use the default constructor or introduce a flag in the EnableBoundingBoxChecks method (e.g., initializeFromSdf = false) to give users control over this behavior. However, I’m concerned that this might add unnecessary complexity and be more confusing than providing an initial, yet meaningful, value upfront.

We should have a test to verify that the values are similar within some reasonable tolerance.

Agreed. I'm including this verification in the next PR I’m about to open to update the physics system. As the physics system will be able to use the SDF API as a fallback in case the engine does not support the AABB feature, I also added 0c79183 where ComputeAxisAlignedBox is made public.