Game Development Community

T3D 1.2 - mObjToWorld doesn't properly store objects rotation

by Lukas Joergensen · in Torque 3D Professional · 07/28/2012 (12:42 pm) · 5 replies

I have brought this up a couple of times, tried to find a logical error or the cause of this.

I have described the problem in these two threads:
http://www.garagegames.com/community/forums/viewthread/130941
http://www.garagegames.com/community/forums/viewthread/131014/1#comment-831344

As you might have seen, when I use my MeshEmitter on skinmeshes, it works fine. But when I use it on static meshes, the rotation is off.

img20.imageshack.us/img20/209/screenshot01600000.png

Now I have researched it thoroughly and it seems to be because when I retrieve the mVertexData from the static meshes it returns the objectspace coordinates. And the mObjToWorld transform matrix does not rotate the vertices properly.

Using alternative rotation functions, I were able to get the proper rotation for the static objects.

This is the proper rotate function:
Point3F* MeshEmitter::rotate(MatrixF trans, Point3F p)
{
	Point3F* r = new Point3F();
	r->x = p.x * trans[0] + p.y * trans[1] + p.z * trans[2];
	r->y = p.x * trans[4] + p.y * trans[5] + p.z * trans[6];
	r->z = p.x * trans[8] + p.y * trans[9] + p.z * trans[10];
	return r;
}

And this is the one for rotating the tower and the one for the plane:
Point3F* MeshEmitter::TSrotate(MatrixF trans, Point3F p)
{
	Point3F* r = new Point3F();
	//Plane
	r->x = p.y * trans[0] + p.z * trans[1] + p.x * trans[2];
	r->y = p.y * trans[4] + p.z * trans[5] + p.x * trans[6];
	r->z = p.y * trans[8] + p.z * trans[9] + p.x * trans[10];
	//Tower
	r->x = p.x * trans[0] + p.z * trans[1] + p.y * trans[2];
	r->y = p.x * trans[4] + p.z * trans[5] + p.y * trans[6];
	r->z = p.x * trans[8] + p.z * trans[9] + p.y * trans[10];
	return r;
}

Which seems right based on how the gismo is turned when turning on objectcoordinates.

Now, I thought that when I would rotate the point with the mObjToWorld matrix it would be transformed from object coordinates to world coordinates, this doesn't seem to be the case.
Maybe I am just handling this completely wrong but the threads I have found on transforming objectspace coordinates to world space coordinates says it should just be rotated via a mulP to convert it from objectspace to worldspace.
NextNext Thread
PrevPrevious Thread
More from Torque 3D Professional


Torque Owner Lukas Joergensen
WinterLeaf Entertainment
#1
07/30/2012 (7:08 am)
Okay so after playing a little around with it I realised that static meshes usually have a root node that contains it's rotation. And if it has more than that one node it should be a skinmesh, I think, anyway I made this little work around based on those assumptions:
trans = (dynamic_cast<TSStatic*>(SB))->getTransform();
						nodetrans = dynamic_cast<TSStatic*>(SB)->getShapeInstance()->mNodeTransforms[0];
						mat.mul(trans, nodetrans);
mat.mulV((*vertPos * dynamic_cast<TSStatic*>(SB)->getScale()),p);
						pNew->pos = dynamic_cast<TSStatic*>(SB)->getPosition() + *p + (vertexList[co].normal() * ejectionOffset);

And it is working properly for my models now. Can't tell if it will work for all models but atleast it's some kind of a fix..

Torque Owner Guy Allard
Default Studio Name
#2
07/30/2012 (10:50 am)
Beat me to it. Every mesh has a root node transform that gets applied to the verts.

Now that you've found that, can we expect to see particle systems for animated meshes too ? :)

Torque Owner Lukas Joergensen
WinterLeaf Entertainment
#3
07/30/2012 (10:55 am)
What do you mean with animated meshes?
It has worked with animated meshes all the time! The culprit was the static meshes!
But now that I've fixed it I am working on documentation and code cleaning and the update will soon go live!
After this final IPS Lite update I will work on the IPS Full version, more info on this is coming soon but I ensure you it will be a greater pack than you can imagine ;)

Torque Owner Guy Allard
Default Studio Name
#4
07/30/2012 (11:04 am)
Ah, ok I didn't realise that you already had the animated meshes working. I must make time to check this out properly, as you seem to be making some great progress with it.

Torque Owner Ivan Mandzhukov
Liman3D
#5
07/30/2012 (11:29 am)
Since you have same axis aligned ortho transforms, it is not necessarily that mObjToWorld to apply rotation,but only translation.