Case Study on Message Races in Data Distribution Service Programs

Data Distribution Service (DDS) is a dependable communication middleware architecture to provide real-time interoperable data exchanges for airborne software. It is important to efficiently detect message races for debugging DDS programs, because it is the most serious type of software faults. Because message races in DDS cause unpredictable results, they must be detected for debugging. However, there has been no report yet on the existence and its confirmation of unintended message races in DDS programs. This paper empirically presents fault cases how it lead to message races on the different order of events. In order to prove it, we show four cases of message races considering the order of two write events and the quality of service (QoS), which lead to nondeterministic results.