Preserving Reactiveness: Understanding and Improving the Debugging Practice of Blocking-Call Bugs

Reactive programming reacts to data items as they occur, rather than waiting for them to complete. This programming paradigm is widely used in asynchronous and event-driven scenarios, such as web applications, microservices, real-time data processing, IoT, interactive UIs, and big data. When done right, it can offer greater responsiveness without extra resource usage. However, this also requires a thorough understanding of asynchronous and non-blocking coding, posing a learning curve for developers new to this style of programming. In this work, we analyze issues reported in reactive applications and explore their corresponding fixes. Our investigation results reveal that (1) developers often do not fix or ignore reactiveness bugs as compared to other bug types, and (2) this tendency is most pronounced for blocking-call bugs -- bugs that block the execution of the program to wait for the operations (typically I/O operations) to finish, wasting CPU and memory resources. To improve the debugging practice of such blocking bugs, we develop a pattern-based proactive program repair technique and obtain 30 patches, which we submit to the developers. In addition, we hypothesize that the low patch acceptance rate for reactiveness bugs is due to the difficulty of assessing the patches. This is in contrast to functionality bugs, where the correctness of the patches can be assessed by running test cases. To assess our hypothesis, we split our patches into two groups: one with performance improvement evidence and the other without. It turns out that the patches are more likely to be accepted when submitted with performance improvement evidence.