Path-oriented Random Testing (PRT) aims at generating a uniformly spread out sequence of random test data that activate a single control flow path within an imperative program. The main challenge of PRT is to build efficiently such a test suite in order to minimize the number of rejects (test data that activate another control flow path). We address this problem with an original technique based on constraint reasoning over finite domains, a well-recognized Constraint Programming technique. Our approach derives path conditions by using symbolic execution and computes an approximation of their associated subdomain by using constraint propagation and constraint refutation.