oh shoot, I'm such an idiot -.-
// when aligned with player's up (first mode, and currently the only one implemented)
// project the player's view forward onto his entity's XZ plane and
// use the closest cardinal direction it's pointing in
- const glm::vec3 view_forward(-GetPlayer().GetEntity().ViewTransform(GetPlayer().GetEntity().ChunkCoords())[3]);
+ const glm::vec3 view_forward(-GetPlayer().GetEntity().ViewTransform(GetPlayer().GetEntity().ChunkCoords())[2]);
// if view is straight up or down, this will be a null vector (NaN after normalization)
// in that case maybe the model forward should be used?
// the current implementation implicitly falls back to TURN_NONE which is -Z