Branch testing of concurrent programs using Petri net models

Most methods for testing sequential programs are useless in testing concurrent programs, owing to the existence of multiple execution flows in concurrent programs. In the paper, a systematic branch testing method for concurrent programs is proposed. Paths generated by the proposed method can be sensitized to obtain the necessary test data. In particular, the proposed method possesses generality, and is capable of detecting all dead codes and deadlocks due to incorrect program design. For evaluating different tests, two new coverage measures used for estimating the quality of testing concurrent programs are proposed. It is shown here that to find optimal paths for branch testing of concurrent programs is NP-hard. The path selection problem with budget limitation is also discussed. Experimental results provide support for the proposed method.