Time-leverage point detection for time sensitive software maintenance

Abstract

Correct real-time behavior is an important aspect for time sensitive software, but it is difficult to get right. Time faults can be introduced not just during software development but also maintenance. So software maintainers without time information tend to have more chances to introduce unintended time behaviors. In this paper, we propose time change impact analysis to help maintainers estimate the potential influence of time changes on programs before the software evolves. Our main insight is that by being reminded and warned that a small-time change at some places in the source code will largely affect the whole task execution time, maintainers can be more cautious when updating such places. Because these places have a leverage effect that multiplies the task execution time in a subtle way, we call them time-leverage points. We give an approach to detect the time-leverage points based on a dynamic testing method, which instruments the program at a point for introducing a small delay and observes its impact on the task execution time. We implement a prototype tool and empirically evaluate the approach.