【摘要】 在不分析库代码方法体的前提下,提出一种支持应用部分局部调用图生成的指针分析方法.该方法通过构建一系列规则,对应用部分和库部分的交互行为进行建模,推导库部分的指针信息.基于Soot程序分析框架实现该方法,并在14个Java基准程序上对其性能以及所生成调用图的完整性和精确性进行评估.实验结果表明:该方法的运行速度比Averroes和Spark调用图生成方法分别快4.9倍和13.7倍,并且能够创建完整且精确的局部调用图.更多还原

【Abstract】 A points-to analysis approach was proposed to generate partial call graphs of application part without analyzing the method bodies of libraries.A set of rules was bulit to model the interaction between application and libraries in order to infer the pointer information libraries.The approach was implemented based on the Soot program analysis framework.The performance,completeness and precision of generated call graphs were evaluated on 14 Java benchmarks.The experimental results showed that the proposed approach was faster than Averroes and Spark call graph construction approaches by a factor of 4.9xand13.7xrespectively.Meanwhile,the proposed approach can construct complete and precise partial call graphs.更多还原