The Failure Behaviors of Multi-Faults Programs: An Empirical Study

Multi-faults contained in a program can manifest themselves as unexpected failure behaviors. Understanding the failure behaviors of multiple faults is essential for enhancing strategies of program debugging, software fault detection and software maintenance. This paper presents an empirical study of the failure behaviors of multi-faults for four industrial software systems. These four software systems contain a total of 128 software faults, which are identified by 3111 test cases. Our observations show that: 1) independent assumption holds true in most cases; 2) less than 3 faults in multi-faults programs are concurrently trigged by one test case; 3) the failure behaviors of multi-faults can mainly be accounted by dominant faults; 4) the interactions of the dominant faults and recessive faults can cause unexpected behaviors of the system.