The newton solver iterates until it gets within a program specified c…

…loseness tolerance to the ray under consideration. The UV parameters are then checked to determine if they are within the current surface subdivision bounding (SSB) box under evaluation. For rays close to the UV border the solver occasionally fails because it jumps slightly out of its current UV but within the ray closeness tolerance. These rays are usually picked up when evaluating the adjacent SSB but not always. To make sure these valid hit points aren't dropped I've added a slight tolerance (VUNITIZE_TOL) to the UV check. This fix is related to speckling picked up on raytraced images of implicit geometries converted to a BREP.

svn:revision:55755
svn:branch:trunk
svn:account:indianlarry

src/librt/primitives/brep/brep.cpp

@@ -788,10 +788,13 @@ utah_newton_solver(const BBNode* sbv, const ON_Surface* surf, const ON_Ray& r, O
 	}
 
 	if (rootdist < ROOT_TOL) {
-	    if (sbv->m_u.Includes(uv.x) && sbv->m_v.Includes(uv.y)) {
+	    int ulow = (sbv->m_u.m_t[0] <= sbv->m_u.m_t[1]) ? 0 : 1;
+	    int vlow = (sbv->m_v.m_t[0] <= sbv->m_v.m_t[1]) ? 0 : 1;
+	    if ((sbv->m_u.m_t[ulow]-VUNITIZE_TOL < uv.x && uv.x < sbv->m_u.m_t[1-ulow]+VUNITIZE_TOL) &&
+		    (sbv->m_v.m_t[vlow]-VUNITIZE_TOL < uv.y && uv.y < sbv->m_v.m_t[1-vlow]-VUNITIZE_TOL)) {
 		bool new_point = true;
 		for (int j=0;j<count;j++) {
-		    if (NEAR_EQUAL(uv.x, ouv[j].x, 0.0001) && NEAR_EQUAL(uv.y, ouv[j].y, 0.0001)) {
+		    if (NEAR_EQUAL(uv.x, ouv[j].x, VUNITIZE_TOL) && NEAR_EQUAL(uv.y, ouv[j].y, VUNITIZE_TOL)) {
 			new_point = false;
 		    }
 		}