BTW guys, I think I have tracked down this bug:
[]
UE-15043 DeprojectScreenPositionToWorld has inaccurate behavior in VR
[/]
AnserHub Bug Report for UE-15043
Here’s PlayerController code for DeprojectScreenPositionToWorld.
Here’s CalcSceneView in LocalPlayer.
It appears problem is caused because CalcSceneView does not get passed info on StereoPass (when called from PlayerController DeprojectScreenPositionToWorld), so it deprojects as if HMD wasn’t connected. Now, I haven’t tested this yet, but my hunch is that properly passing which eye’s view to deproject based on would fix issue.
My question is then what best way to fix this is? Would exposing a parameter for which eye to use to DeprojectScreenPositionToWorld be acceptable? If so I wouldn’t mind testing it and providing a pull request. Otherwise, I’d be happy to submit a PR after some suggestions on proper way to fix.