Btw, some time ago, I was also having issues with INTERSECTION (that created big meshes while tiny were expected), so now I prepared a test case to check if it was still present, but the issue seems gone! Probably in function of the many changes/fixed you provided 
But, while testing with this new code, I went into this exception:
java.lang.IllegalArgumentException: CSGSolid.splitFaces - too many splits:629502
at net.wcomohundro.jme3.csg.iob.CSGSolid.splitFaces(CSGSolid.java:206)
at net.wcomohundro.jme3.csg.iob.CSGShapeIOB.composeSolid(CSGShapeIOB.java:614)
at net.wcomohundro.jme3.csg.iob.CSGShapeIOB.intersection(CSGShapeIOB.java:311)
at net.wcomohundro.jme3.csg.iob.CSGShapeIOB.intersection(CSGShapeIOB.java:1)
at net.wcomohundro.jme3.csg.CSGShape.intersection(CSGShape.java:670)
at net.wcomohundro.jme3.csg.CSGGeometry.regenerate(CSGGeometry.java:398)
at net.wcomohundro.jme3.csg.CSGGeometry.regenerate(CSGGeometry.java:345)
The actual problem with this is that regenerate() may take too long to end with exception. The minimum here took about 30s, but I saw it last like 5min (the most I waited anyway).
So, I would like to suggest this patch at CSGSolid.java (the throw InterruptedException line), but I dont know if it can cause any trouble elsewhere:
public void splitFaces(...){
...
loop1: for( int i = 0, j = initialLimit; i < (j = mFaces.size()); i += 1 ) {
if(Thread.currentThread().isInterrupted())throw new InterruptedException();
With that, we can monitor the thread and interrupt it safely/cleanly at that spot (as just below it has another throw).
The actual point on interrupting is that small changes in the location of shapes may simply allow it to work as expected. Or alternatively, we can skip the regenerate() to release CPU if it is taking too long.
This new (a bit messy) test case below can be run differently by setting bThreaded = true, what demonstrates the InterruptedException âquick recover functionalityâ (if false it will simply pop the splitFaces() exception). Also, it will work without the InterruptedException patch, so the time these threads take to complete, the exceptions and the CPU usage will be noticeable:
public class TestCSGintersection extends SimpleApplication{
public static void main(String[] pArgs) {
SimpleApplication app = new TestCSGintersection();
app.start();
}
private boolean bThreaded = false; // set to true to skip the exceptions and retry
private int iMaxSluggishThreads = Runtime.getRuntime().availableProcessors();
private CSGGeometry geom;
private CSGShape shapePrev;
private CSGShape shapeSphere;
private CSGShape shapeSphereI;
private int iInteractionCount;
private boolean bDancing=false;
ArrayList<RThread> aSluggishThreads = new ArrayList<RThread>();
ArrayList<Vector3f> apos = new ArrayList<Vector3f>();
ArrayList<Vector3f> aposI = new ArrayList<Vector3f>();
private Material mat;
private long ltimep;
private CSGGeometry geomI;
private BoundingBox bbPrev = new BoundingBox();
public TestCSGintersection() {
super( new StatsAppState(), new FlyCamAppState(), new DebugKeysAppState() );
}
@Override
public void simpleInitApp() {
geom = new CSGGeometry("hi");
CSGShape aCube = new CSGShape("Box", new CSGBox(1,1,1));
geom.addShape(aCube);
geom.regenerate();
mat = new Material( assetManager, "Common/MatDefs/Misc/ShowNormals.j3md" );
geom.setMaterial( mat );
shapeSphere = new CSGShape( "Sphere", new CSGSphere( 5, 5, 0.3f ) );
shapeSphereI = new CSGShape( "Sphere", new CSGSphere( 5, 5, 0.3f ) );
rootNode.attachChild( geom );
geomI = new CSGGeometry("hiI");
geomI.setMaterial(mat);
geomI.move(new Vector3f(2,0,0));
rootNode.attachChild(geomI);
testErrSplits();
System.out.println("private void testErrSplits(){");
}
/**
* intersection pos is predicted
*/
private void testErrSplits(){
apos.add(new Vector3f(0.83990216f,0.01416303f,0.41397852f)); //1, [0:1454110976715ms][0:(0.0, 0.0, 0.0);0.0]
apos.add(new Vector3f(0.14809741f,0.30853847f,0.32987046f)); //2, [1:58ms][1:(0.1648531, 0.2834636, 0.3);0.4444414]
apos.add(new Vector3f(0.54515183f,0.29772508f,0.26086247f)); //3, [2:11ms][2:(0.2713526, 0.28531697, 0.3);0.49501315]
apos.add(new Vector3f(0.50555390f,0.28294140f,0.73018849f)); //4, [3:8ms][3:(0.2713526, 0.27231753, 0.3);0.4876362]
apos.add(new Vector3f(0.67798418f,0.78858733f,0.27823582f)); //5, [4:3ms][4:(0.27135253, 0.28531694, 0.29456878);0.4917405]
apos.add(new Vector3f(0.00424497f,0.13861489f,0.73975629f)); //6, [5:5ms][5:(0.23324549, 0.24334115, 0.28711078);0.44277644]
apos.add(new Vector3f(0.37070101f,0.58056056f,0.03278502f)); //7, [6:4ms][6:(0.24722548, 0.28531694, 0.24818724);0.45179987]
//errDelay:29.85s
/*
java.lang.IllegalArgumentException: CSGSolid.splitFaces - too many splits:629502
at net.wcomohundro.jme3.csg.iob.CSGSolid.splitFaces(CSGSolid.java:206)
at net.wcomohundro.jme3.csg.iob.CSGShapeIOB.composeSolid(CSGShapeIOB.java:614)
at net.wcomohundro.jme3.csg.iob.CSGShapeIOB.intersection(CSGShapeIOB.java:311)
at net.wcomohundro.jme3.csg.iob.CSGShapeIOB.intersection(CSGShapeIOB.java:1)
at net.wcomohundro.jme3.csg.CSGShape.intersection(CSGShape.java:670)
at net.wcomohundro.jme3.csg.CSGGeometry.regenerate(CSGGeometry.java:398)
at net.wcomohundro.jme3.csg.CSGGeometry.regenerate(CSGGeometry.java:345)
*/
}
@Override
public void simpleUpdate(float tpf) {
org.lwjgl.input.Mouse.setGrabbed(false);
if(shapePrev==null)shapePrev = new CSGShape("Box", new CSGBox(1,1,1));
geom.addShape(shapePrev);
String fromArray="";
if(apos.size()>0){
shapeSphere.setLocalTranslation(apos.remove(0));
fromArray="A:";
}else{
shapeSphere.setLocalTranslation(geom.getLocalTranslation());
shapeSphere.move(randPos(null).multLocal(0.9f));
}
System.out.println("apos.add(new Vector3f"+dump(shapeSphere.getLocalTranslation())+"); //"
+fromArray+(++iInteractionCount)+", "
+"["+(iInteractionCount-1)+":"+(System.currentTimeMillis()-ltimep)+"ms]"
+"["+(iInteractionCount-1)+":"+bbPrev.getExtent(null)+";"+bbPrev.getExtent(null).length()+"]"
);
ltimep = System.currentTimeMillis();
geom.addShape( shapeSphere, CSGOperator.DIFFERENCE );
try{
shapePrev = geom.regenerate();
if(bThreaded ){
while(!threadedDance()){
System.err.println("//Too many threads "+aSluggishThreads.size()+", unable to start a new."+System.currentTimeMillis()+"...");
Thread.sleep(3000);
}
}else{
intersectionDance();
}
}catch(Exception ex){
System.err.println("//errDelay:"+(System.currentTimeMillis()-ltimep)/1000f+"s");
System.err.println("/*");
ex.printStackTrace();
System.err.println("*/");
exit1();
}
geom.removeAllShapes();
}
private class RThread extends Thread{
RRun r;
public RThread(RRun r,int i){
super(r);
this.r=r;
setName("thread:"+i);
}
public RRun r(){
return r;
}
}
private class RRun implements Runnable{
long lTime=System.currentTimeMillis();
boolean bSuccess=false;
int i;
public RRun(int i){
this.i=i;
}
@Override
public void run() {
try{
intersectionDance();
bSuccess=true;
}catch(Exception ex){
System.err.println("// EXCEPTION: thread:"+i+", "+getLifeTime()+", "+ex.getMessage());
return; //would still be dancing...
}
bDancing=false;
}
public String getLifeTime(){
return ""+(System.currentTimeMillis()-lTime)/1000f+"s";
}
}
private boolean threadedDance(){
boolean bStartedNewThread=false;
// avoid clog cpu
if(aSluggishThreads.size() < iMaxSluggishThreads ){
bDancing=true;
RThread t = new RThread(new RRun(iInteractionCount),iInteractionCount);
t.start();
bStartedNewThread=true;
aSluggishThreads.add(t);
// int iHundredths=10;//100;
int iHundredths=500;
try {
int iSleepCount=0;
while(++iSleepCount<iHundredths){
Thread.sleep(10);
if(!t.isAlive()){
// System.err.println("// THREAD-ENDED "+t.getName());
aSluggishThreads.remove(t); //ended properly and in time
break;
}
}
} catch (InterruptedException e) {
e.printStackTrace();
}
if(bDancing){
System.err.println("// THREAD-INTERRUPTING "+t.getName()+" intersection calc totThreads="+aSluggishThreads.size());
t.setPriority(Thread.MIN_PRIORITY);
t.interrupt();
}
}
for(RThread t2 : aSluggishThreads.toArray(new RThread[0])){
if(!t2.isAlive()){
if(t2.r().bSuccess){
System.err.println("// THREAD actually ended properly: "+t2.getName()+", "+t2.r().getLifeTime());
}//else was the exception log message
aSluggishThreads.remove(t2);
}
}
return bStartedNewThread;
}
private void intersectionDance(){
geomI.removeAllShapes(); //begin cleaning
CSGShape shapePrevToI = new CSGShape("cloneToI",geom.getMesh().deepClone());
geomI.addShape(shapePrevToI);
Vector3f pos = shapeSphere.getLocalTranslation();
shapeSphereI.setLocalTranslation(pos);
shapeSphereI.move(randPos(pos.length()).multLocal(0.1f));
geomI.addShape(shapeSphereI, CSGOperator.INTERSECTION);
CSGShape shapeI = geomI.regenerate(); // <---<<
boundsCheck();
}
private void boundsCheck(){
// geomI.updateGeometricState();
geomI.updateModelBound();
BoundingBox bb = ((BoundingBox)geomI.getModelBound());
bbPrev = bb;
if(bb.getExtent(null).length() > 2f){
System.err.println("//"+bb.toString());
exit1();
}
}
/**
*
* @param fSeed grants predictable randoms for predefined positions array
* @return
*/
private Vector3f randPos(Float fSeed){
Random rnd = fSeed==null ? FastMath.rand : new Random(fSeed.longValue());
//return new Vector3f(rnd.nextFloat()*2f-1f, rnd.nextFloat()*2f-1f, rnd.nextFloat()*2f-1f);
return new Vector3f(rnd.nextFloat(), rnd.nextFloat(), rnd.nextFloat());
}
private String dump(Vector3f pos) {
return String.format("(%01.8ff,%01.8ff,%01.8ff)",pos.x,pos.y,pos.z);
}
private void exit1(){
System.err.println("}");
System.exit(1);
}
}
PS.: oh also another subject, I think GeometryBatchFactory can be helpful with CSGGeonode to provide a single mesh with many materials mixed and overall lightweight.
