A fine-grained data set and analysis of tangling in bug fixing commits

Abstract Context Tangled commits are changes to software that address multiple concerns at once. For researchers interested in bugs, tangled mean they actually study not only but also other irrelevant for the of bugs. Objective We want improve our understanding prevalence tangling and types within bug fixing commits. Methods use a crowd sourcing approach manual labeling validate which contribute fixes each line Each is labeled by four participants. If least three participants agree on same label, we have consensus. Results estimate between 17% 32% all modify source code fix underlying problem. However, when consider production files this ratio increases 66% 87%. find about 11% lines hard label leading active disagreements Due confirmed uncertainty data, 3% 47% data noisy without untangling, depending case. Conclusion high can lead large amount noise data. Prior research indicates may alter results. As researchers, should be skeptics assume unvalidated likely very noisy, until proven otherwise.