A Diagnostic Approach to Test Prioritization

In development processes with high code production ratestesting typically triggers fault diagnosis to localize the de-tected failures. However, current test prioritization algo-rithms are tuned for failure detection rate rather than di-agnostic information. Consequently, unnecessary diagnosticeffort might be spent to localize the faults. We present adynamic test prioritization algorithm that trades fault de-tection rate for diagnostic performance, minimizing overalltesting and diagnosis cost. The algorithm exploits pass/failinformation from each test to select the next test, optimizingthe diagnostic information produced per test. Experimen-tal results from synthetic test suites, and suites taken fromthe Software-artifact Infrastructure Repository show possi-ble diagnostic cost reductions up to 10 and 19 percent, re-spectively, compared to the best of random selection, FEP,and ART. The cost reduction is sensitive to the quality ofthe test coverage matrices and component health, but tendsto grow with the number of faults.