PAF a technique that prioritzes automatically-generated test failures based on the likelihood of violating implicit preconditions.
PAF
(1) clusters generated test failures due to the same cause,
(2) partitions the clusters into two groups, likely fault-revealing and non-fault-revealing, and
(3) ranks the paritioned clusters based on thier likelihood of violating implicit preconditions.
This Github provides the experimental data of PAF and the tool to replicate the results. The results of each subject contain the following components:
test-flow-sets: clusters of tests sharing the same crash origin, crash variable, and crash statement.
FR-cluster-OPT: optimal set of fault revealing clusters that fail due to the same cause. This is used as the ground trugh of RQ1.
FR-cluster-PAF: set of fault revealing clusters returned by PAF. This also refers to fault revealing test flow-sets.
FR-cluster-MSeer, FR-cluster-ReBucket: clusters returned by existing clustering techniques, MSeer and ReBucket, respectively.
ranking_PAF: ranking results of PAF's test-flow-sets.
ranking_PAF_alltests: ranking results of PAF's individual tests.
ranking_others.csv: ranking results of JCrasher and Daikon's.