Null pointer dereference raises Null Pointer Exceptions (NPEs). There are two groups of approaches to detect NPEs. Type-based approaches carry out strict type-based null safety checking. They heavily rely on annotations, and thus produce many false positives. Dataflow-based approaches leverage static forward and/or backward dataflow analysis. They mostly have a limited capability in tracking fields and interprocedural analysis, and introduce false positives and false negatives. To address these drawbacks, we propose Wheeljack to detect NPEs for Java. It does not rely on annotations, and hence can work effectively under a lack of annotations. It leverages our novel abstraction of nullness status to enhance field tracking, and our novel invocation analysis (capturing change to return value and side effect of an invocation) to enhance interprocedural analysis. Our evaluation on 28 Java projects has demonstrated that Wheeljack can mostly outperform the four state-of-the-art NPE detectors in recall without sacrificing precision. 5 and 2 new NPEs have been confirmed and fixed by developers after we submit 8 issues.

Libraries are widely adopted in developing software projects. Library APIs are often missing during library evolution as library developers may deprecate, remove or refactor APIs. As a result, client developers have to manually find replacement APIs for missing APIs when updating library versions in their projects, which is a difficult and expensive software maintenance task. One of the key limitations of the existing automated approaches is that they usually consider the library itself as the single source to find replacement APIs, which heavily limits their accuracy.In this paper, we first present an empirical study to understand characteristics about missing APIs and their replacements. Specifically, we quantify the prevalence of missing APIs, and summarize the knowledge sources where the replacements are found, and the code change and mapping cardinality between missing APIs and their replacements. Then, inspired by the insights from our study, we propose a heuristic-based approach, REPFINDER, to automatically find replacements for missing APIs in library update. We design and combine a set of heuristics to hierarchically search three sources (deprecation message, own library, and external library) for finding replacements. Our evaluation has demonstrated that REPFINDER can find replacement APIs effectively and efficiently, and significantly outperform the state-of-the-art approaches.