Log message: bug3 shows what may or may be not a bug. In any case, Polyhedra::conversion() is entered with a big (but not terrible) polyhedron and control seems not to comes back, while memory usage seems to grow slowly but continuously. It may be "the concrete effects of combinatorial explosion" or a genuine bug.

I have started playing with the example. When considering the first 33 constraints, it returns almost immediately that the polyhedron is not empy; when considering 34 constraints, it returns after some time (say, a couple of minutes on gandalf, which is slow and has few memory); with 35 constraints it is still running after 6 minutes of CPU time. It would be also interesting to experiment with different segments of the matrix, e.g. by using the LAST rows instead of the first ones ... Enea.