chpatton013 / crash Goto Github PK

Two separate problems have been identified

• Final bone transformation matrices contain negative scale members.
  • Other members may also be incorrect.
• Vertex attribute bone weights are all zero on the graphics device.
  • These values appear to be non-zero before being sent to the graphics device.

Event-based collision detection would require tracking which Boundable’s are colliding at each update cycle and only notifying the user of the library when this state changes. Implementing this in Driver would incur a heavy cost for all users, even those that only care to see collision polling reported. A more sustainable solution is to allow users (or State’s) to implement the collision tracking to achieve event-based collision detection.

Objects are clipped from the scene well before they should be in certain scenarios.

A reproducible case has been found that exposes this behavior:

1 Actor
position: <0, 0, 0>
orientation: <0, 1, 0>, 45 degrees
size: <1, 1, 1>

Camera
position: <-1, 0, 1>
orientation: <0, 1, 0>, 45 degrees; <1, 0, 0>, 10 degrees

A State interface would be created that listed onEnter and onExit functions, as well as all of the callback registration currently found in Driver. Driver would have to be modified to operate a current State* member, but the overall implementation of Driver would become drastically simplified.

AssImp allows several textures to exist for each channel. Crash currently uses the first textures found in each of the displacement, normal, ambient, diffuse, specular, and shininess texture channels to populate the graphics device. However, if every texture found in each channel were used, the notion of a skin could be created by mapping each channel to the texture index to use for each MeshInstance.