The PolygonHandlerTest and Rendering2DTest files try to project coordinates very far from the map projection area of validity, leading to results 2 million meters far from the expected values. Those errors were unoticed before because in our current configuration, Maven 1 do not enable J2SE 1.4 assertions when running tests (which is quite unfortunate for a test suite). But our Maven 2 build do enable assertions. Consequently, the Maven 2 build reveals some errors unoticed before. Stack trace below (note: please update to revision 17432 in order to get the same error message. I tried to make it more explicit): java.lang.AssertionError: The transform result may be 2,280,332.971 meters away from the expected position. Are you sure that the input coordinates are inside this map projection area of validity? The point is located 102°30.9'E away from the central meridian and 0°00.0'N away from the latitude of origin. at org.geotools.referencing.operation.projection.MapProjection.checkTransform(MapProjection.java:503) at org.geotools.referencing.operation.projection.MapProjection.transform(MapProjection.java:640) at org.geotools.referencing.operation.projection.MapProjection.transform(MapProjection.java:675) at org.geotools.referencing.operation.transform.ConcatenatedTransformDirect.transform(ConcatenatedTransformDirect.java:81) at org.geotools.renderer.shape.GeometryHandlerUtilities.calculateSpan(GeometryHandlerUtilities.java:101) at org.geotools.renderer.shape.PolygonHandler.<init>(PolygonHandler.java:64) at org.geotools.data.shapefile.shp.PolygonHandlerTest.testRead(PolygonHandlerTest.java:64) As the error message say, PolygonHandler tries to project a point very far from the central meridian (too far for the Transverse Mercator "NAD83 / UTM zone 10N" projection used in this test). I have not checked, but I suspect that it tries to project the (0,0) coordinates or something close from that. If this hypothesis is right, than PolygonHandler should make sure that it transforms only actual coordinates from the polygon, not "empty slot" or "dummy" coordinates. TIP 1: if an empty slot (filled with dummy coordinates) is needed in a float[] or double[] buffer, it should be filled with (NaN,NaN) values instead of (0,0), which is a valid coordinate. (NaN,NaN) coordinates should not throw any exception or assertion failure in the referencing module (if it does, then please let me know and I will fix that bug). TIP 2: in order to test this issue without Maven 2, try to define the following environmental variable and run the test suite with Maven 1. I didn't tried, but it may work. set MAVEN_OPTS=-ea