2015 Symposium Posters

Project Members
Sze Yiu Chau, Hyojeong Lee, Byungchan An, Julian Dolby and Cristina Nita-Rotaru

Many web applications are written in an asynchronous style, in which logic is triggered in response to network and user events. While this approach has performance benefits and can provide improved user experience, it also makes applications more error prone since the most used languages such as HTML and JavaScript do not provide any explicit support for concurrency control. We present MIRROR, a minimally-invasive race detector for client-side web applications which leverages recording and automated replaying of network events. Our tool uses a static approximation of happens-before ordering to automatically generate different testing scenarios by changing the order of these network events. Our tool is browser agnostic and can be used for both debugging and race finding as it does not require repeated interaction with the production server. We evaluate MIRROR using a benchmark of eight applications, where each captures a representative buggy coding pattern. Out of the eight applications, MIRROR was able to manifest and detect the bug for seven of them.