Intelligent Fault Notification through Understanding Developer Behavior

Опубликовано 7 сентября 2016, 17:03

The earlier in the software development lifecycle that a fault is detected, the cheaper it is to fix. Static analysis tools offer a low cost means of detecting potential software faults during code implementation. However, static analysis tools often generate large amounts of output, and many static analysis alerts are ignored by the developer because they are deemed unimportant by developer. Tools that provide inaccurate or unimportant information can become like a constantly ringing alarm clock and will eventually be ignored by the developer. In this talk, I will describe our ongoing research with the MimEc plug-in for the Eclipse development environment. MimEc aims to increase the usage and usability of static analysis tools by `mimicking` developer behavior in deciding which alerts to address and when to address them. In a controlled study, we observed that developers would interrupt their coding task to address a static analysis alert if the alert indicated a potentially critical fault or if it was relevant to the developer`s current working context. We are currently researching when to notify developers of static analysis alerts to maximize the likelihood they will notice and address the alerts.