[11078] Implement a better approximation for vmap liquid levels with …

…diverging "up" directions.

Okay, so game designers CAN fool gravity...I bet they can tilt their cup without spilling the coffee IRL too :p
Only way to do it precisely would be to generate a polygon mesh from the height map and resort to (much slower) raytracing.

src/game/vmap/ModelInstance.cpp

@@ -137,12 +137,14 @@ namespace VMAP
         // child bounds are defined in object space:
         Vector3 pModel = iInvRot * (p - iPos) * iInvScale;
         //Vector3 zDirModel = iInvRot * Vector3(0.f, 0.f, -1.f);
-        float zDist;
-        if (info.hitModel->GetLiquidLevel(pModel, zDist))
+        float zLevel;
+        if (info.hitModel->GetLiquidLevel(pModel, zLevel))
         {
             // calculate world height (zDist in model coords):
-            // assume WMO not tilted (wouldn't make much sense anyway)
-            liqHeight = zDist * iScale + iPos.z;
+            // despite making little sense, there ARE some (slightly) tilted WMOs...
+            // we can only determine liquid height in LOCAL z-direction (heightmap data),
+            // so with increasing tilt, liquid calculation gets increasingly wrong...not my fault, really :p
+            liqHeight = (zLevel - pModel.z) * iScale + p.z;
             return true;
         }
         return false;

src/shared/revision_nr.h

@@ -1,4 +1,4 @@
 #ifndef __REVISION_NR_H__
 #define __REVISION_NR_H__
- #define REVISION_NR "11077"
+ #define REVISION_NR "11078"
 #endif // __REVISION_NR_H__