Towards a better collaboration of static and dynamic analyses for testing concurrent programs

  • Authors:

  • Jun Chen;Steve MacDonald

  • Affiliations:

  • University of Waterloo, Ontario, Canada;University of Waterloo, Ontario, Canada

  • Venue:
  • PADTAD '08 Proceedings of the 6th workshop on Parallel and distributed systems: testing, analysis, and debugging
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Testing concurrent programs remains a difficult task due to the non-deterministic nature of concurrent executions. Many approaches have been proposed to combine static and dynamic analysis to reduce the complexity of uncovering potential concurrency bugs. However, the existing collaboration schemes only provide a limited mechanism for exchanging relevant information between the two analyses. For example, alias information only flows from the static analysis module to the dynamic analysis module at the beginning of the dynamic analysis. Therefore, we cannot fully exploit the advantages of each type of analysis. Motivated by this observation, in this paper we present a new testing technique which enables a tighter collaboration between static analysis and dynamic analysis. In this collaboration scheme, static analysis and dynamic analysis interact iteratively throughout the whole testing process. Static analysis uses coarse-grained analysis to guide the dynamic analysis to concentrate on the relevant search space, while dynamic analysis collects concrete runtime information during the guided exploration. The runtime information provided by the dynamic analysis helps the static analysis to refine its coarse-grained analysis and provides better guidance on dynamic analysis. Currently, our implementation consists of a static analysis module based on Soot and a dynamic analysis module based on JPF (Java PathFinder).