glTF2 Importer uses incorrect code to calculate horizontal FOV

[jamesn031199]

The importer code for glTF2.0 assumes multiplying the vertical FOV by the aspect ratio is how horizontal FOV is calculated.
However, this is not correct and the correct way to convert between the two is by this method referenced in this Reddit post.

[kvnnap]

This code does not respect the aiCamera's interface. It computes the correct horizontal FOV, however, the camera struct aiCamera mHorizontalFOV is defined as being half the horizontal field of view according to the documentation. I have noticed this issue since when I import FBX files the cameras load properly, however with gltf2 the FOV is always wrong. I tested a scene with cameras in this online gltf viewer, and indeed they load fine here. Additionally, I checked the FBX importer code here. This code computes half the horizontal FOV as seen below:

// FBX fov is full-view degrees. We want half-view radians.
out_camera->mHorizontalFOV = AI_DEG_TO_RAD(fov_deg) * 0.5f;

The code here should therefore change from:

aicam->mHorizontalFOV = 2.0f * std::atan(std::tan(cam.cameraProperties.perspective.yfov * 0.5f) * ((aicam->mAspect == 0.f) ? 1.f : aicam->mAspect));

to

aicam->mHorizontalFOV = std::atan(std::tan(cam.cameraProperties.perspective.yfov * 0.5f) * ((aicam->mAspect == 0.f) ? 1.f : aicam->mAspect));

I will gladly provide a pull request if this is agreed to be correct.

Update:

I noticed that the semantics of mHorizontalFOV may be changing from half to full FOV. In which case, this would make the FBX imported not compliant. Not sure what to do.