Accordingly, we extracted 238 explicitly rejected patches from Eclipse and 673 from Mozilla. From these patches, we randomly selected 300 for manual inspection (Table I). Two of the authors individually inspected these 300 rejected patches and particularly their review comments to identify their rejection reasons. Then, the results were compared and similar reasons were merged to one succinct term (e.g., “the patch fails a test” and “test fail” were merged to “test failures”). After this step,the two evaluators agreed on 4/5 patch-rejection reasons, reaching a Cohen’s Kappa of 0.75 that is considered to be “good agreement” [7]. For the remaining 1/5 disagreed reasons, we invited an external computer science postgraduate to the discussion and finally reached an agreement. In total, we manually identified 12 reasons for patch rejection, which are discussed in Section III-A.