Make Quaternion.mult() more safe

Contribution Depot
1

Currently q.mult(q1, q2) allows q1 and q2 as same instance.

if q and q2 is the same instance the operation will fail.



[patch]

Index: src/core/com/jme3/math/Quaternion.java

===================================================================

— src/core/com/jme3/math/Quaternion.java (revision 6541)

+++ src/core/com/jme3/math/Quaternion.java (working copy)

@@ -819,13 +819,15 @@

public Quaternion mult(Quaternion q, Quaternion res) {

if (res == null)

res = new Quaternion();

  •    float qw = q.w, qx = q.x, qy = q.y, qz = q.z;<br />

  •    res.x = x * qw + y * qz - z * qy + w * qx;<br />

  •    res.y = -x * qz + y * qw + z * qx + w * qy;<br />

  •    res.z = x * qy - y * qx + z * qw + w * qz;<br />

  •    res.w = -x * qx - y * qy - z * qz + w * qw;<br />

  •    float qw = q.getW(), qx = q.getX(), qy = q.getY(), qz = q.getZ();<br />

  •    float lw = w, lx = x, ly = y, lz = z;<br />

  •    res.x = lx * qw + ly * qz - lz * qy + lw * qx;<br />

  •    res.y = -lx * qz + ly * qw + lz * qx + lw * qy;<br />

  •    res.z = lx * qy - ly * qx + lz * qw + lw * qz;<br />

  •    res.w = -lx * qx - ly * qy - lz * qz + lw * qw;<br />

return res;

}

+



/**

  • <code>apply</code> multiplies this quaternion by a parameter matrix



    [/patch]
2

looks good.

3

committed in svn. rev 6560

4

There seem to be a problem here.



If I turn my ship left about 90 degrees then pitch up/down the rotation acts as if the ship was facing the same way before it had turned. So, instead of dipping down/up, it’ll bank left/right.



I reverted to 6559 and it works fine after rebuilding everything. Then just to make sure I updated to latest version and the problem is back.



I’m not entirely sure about the next part, but it sure seems like it. I don’t know if it’s only the parameters that are wrongly named, but in spatial.rotation(…), it says:



[java]rotate(float yaw, float roll, float pitch)[/java]



When in fact it’s: pitch, roll, yaw.



Unless I’m wrong and AFAIK, pitch is up/down (dip head), roll is turning left/right (shake head) and yaw is rolling left/right (cocking head).



To ascertain that it wasn’t the model, I re-exported it making sure it was facing the X axis, but the result is the same.



Could anyone confirm please?



Here’s how I do it:

[java]

private AnalogListener analogListener = new AnalogListener() {



@Override

public void onAnalog(String name, float value, float tpf) {

if (name.equals(“PitchDown”)) {

player.getShipSpatial().rotate(.5f * tpf, 0, 0);

galaxyNode.updateGeometricState();

} else if (name.equals(“PitchUp”)) {

player.getShipSpatial().rotate(-.5f * tpf, 0, 0);

galaxyNode.updateGeometricState();

} else if (name.equals(“MoveLeft”)) {

player.getShipSpatial().rotate(0, .5f * tpf, 0);

galaxyNode.updateGeometricState();

} else if (name.equals(“MoveRight”)) {

player.getShipSpatial().rotate(0, -.5f * tpf, 0);

galaxyNode.updateGeometricState();

} else if (name.equals(“RollRight”)) {

player.getShipSpatial().rotate(0, 0, .5f * tpf);

galaxyNode.updateGeometricState();

} else if (name.equals(“RollLeft”)) {

player.getShipSpatial().rotate(0, 0, -.5f * tpf);

galaxyNode.updateGeometricState();

}

}

};

[/java]

5

Yeah, I have strange issues in physics with rotation too since some time… It seems I get zero Matrixes out of some Quaternions lately…

6

reverted the change. rev. 6790

I don’t find any logical error on this patch.

But I reverted because of the wierd errors. rev. 6689, 6616, 6560

7

I confirm that the reverts fix the problem. The ship now behaves the way it was before.