What does this error mean?

[FSA]

When i set a matrix with NewtonBodySetMatrix (RotationMatrix) I get an error:
dgBody.h Line 738
Expression: dgAbsf(val-1.0f) < 1.0e-5f
What does it mean?

[JoeJ]

Either the axis you give are of nonuniform length, or the axis ar not perpendicular to each other.
Mostly this happens because nonuniform scale, see viewtopic.php?f=9&t=7334&start=15 for too much details about that

[FSA]

Oh. You're right...
That's my matrix
-1.#IO -1.#IO -1.#IO 0.000
-1.#IO -1.#IO -1.#IO 0.000
-1.#IO -1.#IO -1.#IO 0.000
0.000 0.000 0.000 1.000
I will chek where the error occurs...

[FSA]

Ok. My code to calculate the RotationMatrix:

Code: Select all

Vector3 m_vYAxis, m_vZAxis, m_vXAxis;
   m_vYAxis = Vector3(0.0f,1.0f,0.0f);
   m_vZAxis = Vector3(0.0f, 0.0f, 1.0f);
   m_vXAxis = Vector3(1.0f, 0.0f, 0.0f);

   // Rotation um die x-Achse
   Matrix mRotation( MatrixRotationX( rot.x ) );
   m_vYAxis = Vector3TransformNormal( m_vYAxis, mRotation );
   m_vZAxis = Vector3Cross( m_vXAxis, m_vYAxis );

   // Rotation um die y-Achse
   mRotation = MatrixRotationY( rot.y );
   m_vXAxis = Vector3TransformNormal( m_vXAxis, mRotation );
   m_vZAxis = Vector3Cross( m_vXAxis, m_vYAxis );

   // Rotation um die z-Achse
   mRotation = MatrixRotationZ( rot.z );
   m_vXAxis = Vector3TransformNormal( m_vXAxis, mRotation );
   m_vYAxis = Vector3TransformNormal( m_vYAxis, mRotation );

   return MatrixAxes( Vector3Normalize(m_vXAxis), Vector3Normalize(m_vYAxis), Vector3Normalize(m_vZAxis) );

Vector3TransformNormal:

Code: Select all

FF_API Vector3 Vector3TransformNormal( const Vector3& v,
                                    const Matrix& m )
{
   const float fLength = Vector3Length( v );
   if( fLength == 0.0f ) return v;

   const Matrix mTransform( MatrixTranspose( MatrixInvert( m ) ) );

   return Vector3Normalize( Vector3( v.x * mTransform.m11 + v.y * mTransform.m21 + v.z * mTransform.m31,
                                      v.x * mTransform.m12 + v.y * mTransform.m22 + v.z * mTransform.m32,
                              v.x * mTransform.m13 + v.y * mTransform.m23 + v.z * mTransform.m33 ) )
         * fLength;
}


MatrixRotationX/Y/Z create a Rotation Matrix. Vectro3Cross is the crossproduct.

Newton Crash when I change the Z Rotation.
Matrix:
0.997 0.000 0.075 0.000
0.000 0.989 0.149 0.000
-0.074 -0.149 0.986 0.000
0.000 0.000 0.000 1.000

Tanks

[JoeJ]

So what you want is to build a MAtrix from Euler Angles?
I assume your code produces a right handed instead of a left handed coordinate system (or vice versa) - or theres another error in it.
At least it's clear that it's very inefficient.

Try to replace it with code from here:
viewtopic.php?f=9&t=7237&p=49723&hilit=euler+angles#p49723
The order seems to be 0x012 for your case.

If you work with public engine (irrlicht ?), can you tell which one and what order is right?
At lot of people have trouble with that issue.

[FSA]

I have let this problem be a problem
I have a new problem(with core 300).
In the picture you can see it. It happens when i pick a body and turn around with the body. To pick a body I add force and torque to the body. What is the problem? It hasn't happened with core 200.

[Julio Jerez]

that a really bad bug that should never ever happens. what are you doing to get that?

It means a body was at rest but somehow after it was deemed to be resting somehow places in the simulation.
I have never seem it happens?

[FSA]

Here a video: https://www.dropbox.com/s/t648v0gn3jyusni/Error.avi
I picked a body, walk with it, put it on an other body an then the error occurs.

2nd question: Is the friction set by default so high that bodys "glue" together? Video: https://www.dropbox.com/s/ijw3v9m821otbus/Record2.avi

And somtimes i get these errors(pictures). Next post the other Picture.

[FSA]

Next picture.

Here is my code:

Code: Select all

/*Callback*/
void BodyCallback (const NewtonBody* body, dFloat timestep, int threadIndex)
{
   NewtonBodyAddForce (body, g_vLinear);
   NewtonBodyAddTorque(body, g_vTorque);
}


I calculate force an torque with formula from JoeJ. I think it is the right calculation because with core 200 it worked.

[Julio Jerez]

No any of those error should be happening, something is very wrong. I you get an lru pout of sync then not else will work.
island row coun will fal because it will add more bodies than were not support to be consired. I cannot imagine how that happened, It is not possible unless you run the engine straight for over 200 years

when you say "I pick a body and walk with it" from where are you doing that? what ever it is is not right.

an not the friction in the get is 0.6 like is has always being the defualt in all version of the engine. I do not know what could that be.

[FSA]

I give you my code to calculate the force I add with NewtonBodyAddForce.

Code: Select all

if(m_pBody)
   {
      if(m_bFirstFrame)
      {
         m_bFirstFrame = false;
         m_mOffsetMatrix = MatrixIdentity();
         NewtonBodyGetMatrix(m_pBody, m_mOffsetMatrix);
         m_fCameraOffsetX = CamHandler->GetCameraAngleX();
      }

      // Body modification
      NewtonBodySetForceAndTorqueCallback(m_pBody, BodyCallback);
      NewtonBodySetContinuousCollisionMode(m_pBody, 1);

      //******************************
      // Translation
      Vector3      vVelocity;
      Vector3      vBodyPosition;
      Vector3      vCameraDir;
      Vector3      vCameraPos;
      Matrix      mMatrix;
      float      fMass, fA, fB, fC;

      NewtonBodyGetVelocity   (m_pBody, vVelocity);
      NewtonBodyGetMatrix      (m_pBody, mMatrix);
      NewtonBodyGetMassMatrix   (m_pBody, &fMass, &fA, &fB, &fC);
      vBodyPosition =   MatrixGetPosition(mMatrix);
      vCameraDir = CamHandler->GetCameraDir();
      vCameraPos = CamHandler->GetCameraPos();
      vCameraPos += vCameraDir * 25.0f;

      m_vLinear = ( (vCameraPos - vBodyPosition) * 20.0f - vVelocity ) * (fMass / fTime);


m_vLinear is the force i apply.(NewtonBodyAddForce (body, m_vLinear);

With "I move with it" I mean, when I walk the body is always on the same distance(see code).

Thank you for reading this and be so patient

[Julio Jerez]

maybe set continue collsion have a bug, try not using that see if it work.

[FSA]

Yes that solved the problem! The problem with the high friction is also solved! So continues collision mode has a bug?

[Julio Jerez]

Ha, so there must be a bug in CCD that is changes the LRU sincronization between the Collision system and the dynamics system.

can you send me a demo that I can try, one that I can turn CCD on and off?

[FSA]

Yes give me a moment...