User mesh doesn't work

[PJani]

I don't get why but usermesh collision doesn't work as it should. Collide callback is beeing called but object just passes thru. I realy dont know what i am missing.
parameters

Code: Select all

   struct QuadFaceIndexStruct
   {
      int indices[4];
      int material_index;
      int normal_index;
      int normals[4];
      int diag_optimiz;
   };

   static float sajz = 16.0f;
   static int test_count[1] = {4};
   static QuadiFaceIndexStruct[1] test_index = {0,1,2,3,0,4,4,4,4,4,22};
   static Ogre::Vector3[5] test_vertex = {Ogre::Vector3(0,1,0),Ogre::Vector3(sajz,1,0),Ogre::Vector3(sajz,1,sajz),Ogre::Vector3(0,1,sajz),Ogre::Vector3(0,1,0)};
   


This is create for usermesh collision.

Code: Select all

      dfVec3 vec_max(sajz,sajz,sajz);
      dfVec3 vec_min(0,0,0);
      
      NewtonCollision* col = NewtonCreateUserMeshCollision(world, &vec_min.x, &vec_max.x, this
                           , _usermeshCollideCallback
                           , _usermeshRayHitCallback
                           , _usermeshDestroyCallback
                           , _usermeshGetCollisionInfo
                           , _usermeshAABBTest
                           , _usermeshGetFacesInAABB
                           , _usermeshSerialize, 0);


      m_body = new OgreNewt::Body(W, col, 0, position);
      m_body->setUserData(new PhysicsUserDataProxy(m_body, this, this));

      NewtonDestroyCollision(col);


usermesh collide callback

Code: Select all

   void TerrainPhysicalChunk::_collisionCollideCallback(NewtonUserMeshCollisionCollideDesc* const collideDescData)
   {
      collideDescData->m_faceCount = 1;
      collideDescData->m_vertexStrideInBytes = sizeof(Ogre::Vector3);
      
      collideDescData->m_faceIndexCount = (int*)test_count;
      collideDescData->m_faceVertexIndex = (int*)test_index;
      
      collideDescData->m_vertex = (float*)(test_vertex[0].ptr());
   }


[Julio Jerez]

Maybe this is wrong
static Ogre::Vector3[5] test_vertex = {Ogre::Vector3(0,1,0),Ogre::Vector3(sajz,1,0),Ogre::Vector3(sajz,1,sajz),Ogre::Vector3(0,1,sajz),Ogre::Vector3(0,1,0)};

in the polygon is a ground level it should be
static Ogre::Vector3[5] test_vertex = {Ogre::Vector3(0,0,0),Ogre::Vector3(sajz,0,0),Ogre::Vector3(sajz,0,sajz),Ogre::Vector3(0,0,sajz),Ogre::Vector3(0,1,0)};

[PJani]

Nop i tried with that no difference.

[Julio Jerez]

maybe the windin is wrong, try inverting it. I beleiev newton and Ogree use inverted system.

My guess is tha the normal should face down, and the winding hould be inverted.

[PJani]

Ok i just stumbled on something REALLY weird... I was debugging and stepping thru the code...

function in coreLibrary_300\source\physics\dgCollisionUserMesh.cpp

Code: Select all

void dgCollisionUserMesh::GetCollidingFaces (dgPolygonMeshDesc* const data) const


line 161

Code: Select all

const dgVector& faceNormal = vertex[normalIndex * stride];


when newton tries to do
const dgVector& faceNormal = vertex[4*3];

which should yield dgVector(0,1,0) yields zero vector

this part gets called

Code: Select all

   DG_INLINE dgVector (const dgFloat32 a)
      : m_type(_mm_set_ps1(a))
   {
   }


instead of(i think)

Code: Select all

   DG_INLINE dgVector (const dgFloat32* const ptr)
      :m_type(_mm_loadu_ps (ptr))
   {
      m_type = _mm_and_ps (m_type, m_triplexMask.m_type);
   }


this solution should work

Code: Select all

const dgVector& faceNormal = vertex+normalIndex * stride;


now the faceNormal is correct, but still not working

the sandbox demo is crashing for me

[Julio Jerez]

line 161

Code: Select all

const dgVector& faceNormal = vertex[normalIndex * stride];


yes that was wrong

it should be this

Code: Select all

         dgVector faceNormal (&vertex[normalIndex * stride]);

I fixed now, please sync and try again, the sandbox demo should no crash

[Julio Jerez]

I found another bug, the vertex count each face was not save when a face was rejected, please sync try again

I also added the debug display callback, you click show debug line, you will see how it should look like

[PJani]

Hmm it still doesnt work...sandbox works fine

O and i use boxes for testing!

I really don't know what i am missing

[Julio Jerez]

try a box without scale, see if that work at all.

[PJani]

Nop that doesnt work eather.

It really doesn't make any sense

[Julio Jerez]

I do not know what could be, other that debug the code myself.

[PJani]

I will try to find the problem my self. I don't want to give you more work than you have
What could cause this to return 0?(Its located in dgNarrowPhaseCollision.cpp line 1917)

Code: Select all

dgInt32 count1 = polygon.CalculateContactToConvexHullDescrete (proxy, scale, invScale);


[Julio Jerez]

that function will return zero if a conve polygon a convex hul shape do not intersect.


if you can put that demo with teh other demo, I cna check it out. make sure you jusrt put a simple demo, so that I do no have to test too much.
I am curious as to why you are not getting it to work, it appear that the Layla on the otehr thread have it working

[PJani]

Ok somehow it started working(don't ask me why 0_0).

This is how it looks...

Code: Select all

19:50:34: m_vertexes: Vector4(0, 0, 0, 1) Vector4(10, 0, 0, 1) Vector4(10, 0, 10, 1) Vector4(0, 0, 10, 1) Vector4(0, 1, 0, 1)
19:50:34: m_facecount:2
19:50:34:    m_vertexfacecount:3
19:50:34: indeces: 3 1 0 0 4 4 4 4 0
19:50:34:    m_vertexfacecount:3
19:50:34: indeces: 3 2 1 0 4 4 4 4 0


and then this happens...(compare the first and second indices) The second face just becomes the first!

Code: Select all

19:50:34: m_vertexes: Vector4(0, 0, 0, 1) Vector4(10, 0, 0, 1) Vector4(10, 0, 10, 1) Vector4(0, 0, 10, 1) Vector4(0, 1, 0, 1)
19:50:34: m_facecount:2
19:50:34:    m_vertexfacecount:3
19:50:34: indeces: 3 2 1 0 4 4 4 4 0
19:50:34:    m_vertexfacecount:3
19:50:34: indeces: 3 2 1 0 4 4 4 4 0


I don't do anything to the verticles or the indeces via code just initalize them. By any chance newton somewhere changes/copies them?

[Julio Jerez]

Code: Select all

19:50:34: m_vertexes: Vector4(0, 0, 0, 1) Vector4(10, 0, 0, 1) Vector4(10, 0, 10, 1) Vector4(0, 0, 10, 1) Vector4(0, 1, 0, 1)
19:50:34: m_facecount:2
19:50:34:    m_vertexfacecount:3
19:50:34: indeces: 3 2 1 0 4 4 4 4 0
19:50:34:    m_vertexfacecount:3
19:50:34: indeces: 3 2 1 0 4 4 4 4 0


That looks like there is still a bug in newton, it seems the vertex index array is point alway to the same face.
I will modify the demo to pass two triangles and see if I can recreate the same bug.
what line of code did you get that from.