Analyzing the Impact of Change in Multi-threaded Programs

Krishnendu Chaterjee, Luca de Alfaro, Vishwanath Raman, and César Sánchez
We introduce a technique for debugging multi-threaded C programs and analyzing the impact of source code changes, and its implementation in the prototype tool \Direct. Our approach uses a combination of source code instrumentation and runtime management. The source code along with a test harness is instrumented to monitor Operating System (OS) and user defined function calls. \Direct tracks all concurrency control primitives and, optionally, data from the program.

\Direct maintains an abstract global state that combines information from every thread, including the sequence of function calls and concurrency primitives executed. The runtime manager can insert delays, provoking thread interleavings that may exhibit bugs that may be difficult to reach otherwise. The runtime manager collects an approximation of the reachable state space and uses this approximation to assess the impact of change in a new version of the program.

In Proceedings of the 13th International Conference on Fundamental Approaches to Software Engineering (FASE'2010), vol. 6013 of Lecture Notes in Computer Science, pp293-307. Springer-Verlag, 2010.
Download: BIB PDF