Abstract
Open source software development enabled distributed teams of programmers to contribute to large software systems that became standards in the operation of government and business. Crowdsourcing went further by enabling contributions in the form of small and independent tasks. This allowed teams to scale from dozens to hundreds of people. While crowdsourcing established as industry practice in the areas of software testing, it is challenging for source code related tasks, e.g., software debugging. One of the reasons is that the complex dependencies in the source code can make many tasks difficult to partition and sequence, and later aggregate their outcomes. I am investigating these problems in the context of failure resolution tasks. A failure resolution task consists of inspecting the source code with the objective to identify and explain the root-cause of a software failure. My approach partitions code inspection into questions that are automatically instantiated from templates. I present here my research plan and the early results of experiments on the efficacy, efficiency, and scalability of my approach.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call