Bullet wrapper issue

[eadle]

The Bullet wrapper currently doesn't allow a btPairCachingGhostObject to getAllContactManfolds. I've tracked this bug down to how a btBroadphasePair temporary instance is stored when trying to access an index in btBroadphasePairArray. It's an easy fix by storing the temporary btBroadphasePair instance in Java instead of C++. This doesn't change how the API is exposed and it looks like this is the desired behaviour anyway (see btBroadphasePair.java).

    /** Temporary instance, use by native methods that return a btBroadphasePair instance */
    protected final static btBroadphasePair temp = new btBroadphasePair(0, false);
    public static btBroadphasePair internalTemp(long cPtr, boolean own) {
        temp.reset(cPtr, own);
        return temp;
    }

I'll add to this tomorrow with an example to reproduce the results, but the following three file changes fixes the issue (this guy had a similar issue). I didn't make a pull request because I'm having no luck with swig. Any tips would be appreciated.

jni/swig-src/collision/collision_wrap.cpp:

SWIGEXPORT jlong JNICALL Java_com_badlogic_gdx_physics_bullet_collision_CollisionJNI_btBroadphasePairArray_1at(JNIEnv *jenv, jclass jcls, jlong jarg1, jobject jarg1_, jint jarg2) {
  jlong jresult = 0 ;
  btAlignedObjectArray< btBroadphasePair > *arg1 = (btAlignedObjectArray< btBroadphasePair > *) 0 ;
  btBroadphasePair *result = 0 ;
  int arg2 ;

  (void)jenv;
  (void)jcls;
  (void)jarg1_;
  arg1 = *(btAlignedObjectArray< btBroadphasePair > **)&jarg1;
  arg2 = (int)jarg2;
  result = (btBroadphasePair *) &((btAlignedObjectArray< btBroadphasePair > const *)arg1)->at(arg2);
  *(btBroadphasePair **)&jresult = result;
  return jresult;
}

jni/swig-src/collision/com/badlogic/gdx/physics/bullet/collision/btBroadphaseArray.java:

public btBroadphasePair at(int n) {
  return btBroadphasePair.internalTemp(CollisionJNI.btBroadphasePairArray_at(swigCPtr, this, n), false);
}

jni/swig-src/collision/CollisionJNI.java:

public final static native long btBroadphasePairArray_at(long jarg1, btBroadphasePairArray jarg1_, int jarg2);

[xoppa]

Thanks, would you mind to include a test to show the issue. Preferable something we can add to the tests.

[eadle]

Sorry for the delay. Finally figured out how to setup the libgdx tests in Android Studio. I think I've come up with a cool test that can only be solved with btPairCachingGhostObject, but it could take a little time before I make a pull request. If you just want a quick test just to reproduce the issue, let me know.

[badlogic]

Any update on this @eadle ?