Minimal raycast test won't work.

[FunkyJive]

Not really sure what is going on here but I am trying to integrate newton into my current project for raycast selecting purposes (I don't need it for physics simulation). I tried to set up a minimal hello world raycast example but not getting any results. Any ideas of what I am doing wrong? Thanks in advance!

Here is the callback function. It is never printing the message and never getting to this callback.

Code: Select all

dFloat FilterRayCast(const NewtonBody* body, const dFloat* normal, int collisionid, void* userdata, dFloat intersetparam)
{
    printf("Hello world you hit me!\n");
    return intersetparam;
}


Here is the main function.

Code: Select all

NewtonWorld * world = NewtonCreate();
    float min[3];
    float max[3];
    for(int i=0; i<3; i++)
    {
        min[i] = -1000;
        max[i] = 1000;
    }

    NewtonSetWorldSize(world, min, max);
    NewtonCollision * col = NewtonCreateBox(world, 10.0, 10.0, 10.0, 1, NULL);
    NewtonBody * body = NewtonCreateBody(world, col);

    float point1[3];
    point1[0] = 50.0;
    point1[1] = 0.0;
    point1[2] = 0.0;
    float point2[3];
    point2[0] = -50.0;
    point2[1] = 0.0;
    point2[2] = 0.0;
    printf("Shooting the ray...");
    NewtonWorldRayCast(world, point1, point2, FilterRayCast, NULL, NULL);


I am using the latest windows version of newton.

Thanks again.

[Julio Jerez]

yes I beleive there is a problem there, I have let that bug go for too long, it is time to fix it.
good you provided such simple demo, I will verify tonight
Thank you for the test.

[FunkyJive]

Glad I can help but this seems to be puzzling me also. I had used a previous version of Newton in a previous game and the raycasting worked perfect. So for this game, I got my old code and just did some updating on the function calls to meet the current requirements and it doesn't work now. I figured I forgot to copy something so I stripped the program down to this bare bones example and still no luck. Let me take a look back at what version I was using when the ray casting was working.

EDIT*

If it is any help here is the header of the version of newton I was using when it did work.

#ifndef NEWTON_H__736495_6495_076__INCLUDED_
#define NEWTON_H__736495_6495_076__INCLUDED_


#define NEWTON_MAJOR_VERSION 2
#define NEWTON_MINOR_VERSION 15

[Julio Jerez]

The engine goes over optimization and new features in each new release.
for the most part I try to keep everything working bu some time a new optimization may create some side effect and break some funtionality.
Specially now that I changed the file format and many SDK demos are disabled, which prevent me from running the smoke test in each realease.

I am gessing I provably have some part commented out while working on something else and I forgot to re-enable it.
not to worry it should be very eassy to see what is wrong.

[Julio Jerez]

Ok there was a bog in the ray cast, but it was not a raycast failure, it was that ray castion was reportion the wrong collision ID whne ray cation a shape.

This si the ray cat rotine for cation and collision Box in local space

Code: Select all

dgFloat32 dgCollisionBox::RayCast (
   const dgVector& localP0,
   const dgVector& localP1,
   dgContactPoint& contactOut,
   OnRayPrecastAction preFilter,
   const dgBody* body,   
   void* userData) const
{

   if (PREFILTER_RAYCAST (preFilter, body, this, userData)) {
      return dgFloat32 (1.2f);
   }


   dgInt32 index = 0;
   dgFloat32 signDir = dgFloat32 (0.0f);
   dgFloat32 tmin = dgFloat32 (0.0f);
   dgFloat32 tmax = dgFloat32 (1.0f);
   for (dgInt32 i = 0; i < 3; i++) {
      dgFloat32 dp = localP1[i] - localP0[i];
      if (dgAbsf (dp) < dgFloat32 (1.0e-8f)) {
         if (localP0[i] <= m_size[1][i] || localP0[i] >= m_size[0][i]) {
            return dgFloat32 (1.2f);
         }
      } else {
         dp = dgFloat32 (1.0f) / dp;
         dgFloat32 t1 = (m_size[1][i] - localP0[i]) * dp;
         dgFloat32 t2 = (m_size[0][i] - localP0[i]) * dp;

         dgFloat32 sign = dgFloat32 (-1.0f);
         if (t1 > t2) {
            sign = 1;
            Swap(t1, t2);
         }
         if (t1 > tmin) {
            signDir = sign;
            index = i;
            tmin = t1;
         }
         if (t2 < tmax) {
            tmax = t2;
         }
         if (tmin > tmax) {
            return dgFloat32 (1.2f);
         }
      }
   }

   if (tmin >= dgFloat32 (0.0f)) {
      _ASSERTE (tmin < 1.0f);
      contactOut.m_normal = dgVector (dgFloat32 (0.0f), dgFloat32 (0.0f), dgFloat32 (0.0f), dgFloat32 (0.0f));
      contactOut.m_normal[index] = signDir;
      //contactOut.m_userId = SetUserData(); // Julio thsi was the Bug wrong collision ID report to callback
      contactOut.m_userId = SetUserDataID();
   } else {
      tmin = dgFloat32 (1.2f);
   }
   return tmin;
}

In the end if the function found a Hit, it was setting the contact normal to the hit normal but the sphere user data pointer.
Long time ago, there user data of a collision shape use to store the collision ID of the shape, but when I added the scene collision, there is a new object called body Proxy, which act as a pseudo body.
The object needs to store information in the collision shape, so I separate the ID and the collsionUserData.
I made the changed on all collision routine, by I forgot to change he raycast routines. So the where reportion a userdata pointer as collision id, usually zero instead for the collisionID.
It is fix now.
The callback now reports

normal,3 { 1.0000000 0.00000000 0.00000000}
intersetparam 0.44999999 float
collisionid 1 int

Get SDK 2.25 and try again please

[FunkyJive]

Thank you so much for the fix! Just downloaded the new version and everything seems to be working fine now. Thanks again for the quick reply and the fix.

[Aphex]

Smashin, cheers for fix