Cues for scent intensification in debugging

Information foraging is a theory to understand how people search for information. In this theory, information scent is the perceived likelihood by the “predator” that a cue will lead to a “prey”. The better the cues, the better the information scent. In automatic debugging, it is the perceived likelihood that the diagnostic report leads to the cause of failures. In this paper, we detail a visualization, offered by the GZoltar toolset, that has the potential to provide better cues. With better we mean providing more information that leads to the fault than, e.g., the source code and code coverage information. The toolset provides a graphical display of the diagnostic reports yielded by well-known debugging techniques. From an information foraging point of view, we argue that the visualization is of added value while debugging. Finally, we report a user study to confirm that GZOLTAR's visualization provides better cues for pinpointing faults.