Abstract
The exception-handling mechanism has been widely adopted to deal with exception conditions that may arise during program executions. To produce high-quality programs, developers are expected to handle these exception conditions and take necessary recovery or resource-releasing actions. Failing to handle these exception conditions can lead to not only performance degradation, but also critical issues. Developers can write formal specifications to capture expected exception-handling behavior, and then apply tools to automatically analyze program code for detecting specification violations. However, in practice, developers rarely write formal specifications. To address this issue, mining techniques have been used to mine common exception-handling behavior out of program code. In this paper, we discuss challenges and achievements in precisely specifying and mining formal exception-handling specifications, as tackled by our previous work. Our key insight is that expected exception-handling behavior may be “conditional” or may need to accommodate “exceptional” cases.