I'm trying to debianize this package, and have encountered a segmentatoin fault.
When I build and test on an amd64 system, everything goes well.
However, when it comes on an i386 system (both on debian jessie and sid),
MolecularSurfaceTest::buckminsterfullerene() receives a segmentation fault.
I've managed to figure out that this segmentation fault is caused by -2 in the Tetrahedron::neighbors,
which seems to mean that DelaunayTriangulation::insertPoint cannot find a neighboring tetrahedron for a new tetrahedron.
Strangely enough, it does not receive a segfalt when it run under valgrind, yet valgrind reports no errors.
I'm not sure why and under what condition it happens, but I'm afraid that chemkit::geometry::planeOrientation goes unstable when it handles near-coplanar points in some environment.
Hi.
I'm trying to debianize this package, and have encountered a segmentatoin fault. When I build and test on an amd64 system, everything goes well. However, when it comes on an i386 system (both on debian jessie and sid),
MolecularSurfaceTest::buckminsterfullerene()receives a segmentation fault. I've managed to figure out that this segmentation fault is caused by-2in theTetrahedron::neighbors, which seems to mean thatDelaunayTriangulation::insertPointcannot find a neighboring tetrahedron for a new tetrahedron. Strangely enough, it does not receive a segfalt when it run under valgrind, yet valgrind reports no errors.I'm not sure why and under what condition it happens, but I'm afraid that
chemkit::geometry::planeOrientationgoes unstable when it handles near-coplanar points in some environment.What do you think?