Generating run-time progress reports for a points-to analysis in Eclipse

Abstract

Eclipse plug-ins have access to a rich collection of GUI components. One such component is the progress bar, which allows a long-running task to provide Eclipse users with feedback about its progress. This paper considers the problem of providing precise progress bar reports for plug-ins that perform static code analysis. Since static analyses often take a long time to execute, progress indicators can inform the user whether the analysis is actually making progress, and how long it will take to complete. This type of responsiveness is essential for providing positive user experience.In this paper we consider points-to analysis, which is a popular static analysis for object-oriented software. Reporting the run-time progress of a points-to analysis requires heuristics for a priori estimates of the total running time of the analysis. We define several such heuristics for a whole-program subset-based points-to analysis for Java, implemented as part of the Soot Eclipse plug-in. We also present an experimental evaluation of the heuristics on a large set of Java programs. These results provide useful insights for the creators of points-to analyses and other static analyses that will be built and distributed in Eclipse.