Scalable Provenance Generation from Flow-Insensitive Points-To Information

Stepan Sindelar, Paddy Krishnan, Bernhard Scholz, K. R. Raghavendra and Yi Lu
Oracle Labs Australia

Points-to analysis is used extensively in the static analysis of object-oriented
programs especially to detect various types of defects. The usual points-to
analysis does not store the justification (or provenance) for the presence of a
tuple in the points-to result. However, a developer using the results of the
points-to analysis requires the justification in order to construct possible
program traces that exhibit the defect. In this presentation, we will outline
the challenges of storing/computing provenance for points-to information over
large code-bases. We will also describe a client-driven solution and its
implementation based on a post-analysis that uses the results of the points-to
analysis.