A tool for testing of parallel and distributed programs in message-passing environments

Due to the non-deterministic behavior of some parallel and distributed programs, addressing the problem of testing in such context is a non-trivial task. A proficient testing phase must be done using tools which record information about a single execution, and which are able to force a concurrent program to exercise a given execution. This paper describes a tool for testing of programs based on the Horus system. Our approach is novel in dealing with events connected to process group handling and group communication. Moreover, we show that using the capabilities of the Horus run-time system can be really helpful in solving testing problems. Our approach has the advantages of requiring no modifications neither in the program nor in the operating system, and to avoid a centralized solution through the use of partial orders of synchronization events.