Proceedings of the 2006 workshop on Parallel and distributed systems: testing and debugging
The Complexity of Predicting Atomicity Violations
TACAS '09 Proceedings of the 15th International Conference on Tools and Algorithms for the Construction and Analysis of Systems: Held as Part of the Joint European Conferences on Theory and Practice of Software, ETAPS 2009,
Meta-analysis for Atomicity Violations under Nested Locking
CAV '09 Proceedings of the 21st International Conference on Computer Aided Verification
A race-detection and flipping algorithm for automated testing of multi-threaded programs
HVC'06 Proceedings of the 2nd international Haifa verification conference on Hardware and software, verification and testing
Sound predictive race detection in polynomial time
POPL '12 Proceedings of the 39th annual ACM SIGPLAN-SIGACT symposium on Principles of programming languages
FM'06 Proceedings of the 14th international conference on Formal Methods
| Hi-index | 0.00 |
We present an automated and configurable technique for runtime safety analysis of multithreaded programs that is able to predict safety violations from successful executions. Based on a formal specification of safety properties provided by a user, our technique enables us to automatically instrument a given program and create an observer so that the program emits relevant state update events to the observer and the observer checks these updates against the safety specification. The events are stamped with dynamic vector clocks, enabling the observer to infer a causal partial order on the state updates. All event traces that are consistent with this partial order, including the actual execution trace, are then analyzed online and in parallel. A warning is issued whenever one of these potential traces violates the specification. Our technique is scalable and can provide better coverage than conventional testing, but its coverage need not be exhaustive. In fact, one can trade off scalability and comprehensiveness: a window in the state space may be specified allowing the observer to infer some of the more likely runs; if the size of the window is 1, then only the actual execution trace is analyzed, as is the case in conventional testing; if the size of the window is ∞, then all the execution traces consistent with the actual execution trace are analyzed.