Re-computing Coverage Information to Assist Regression Testing

Abstract

This paper presents a technique that leverages an existing regression test-selection algorithm to compute accurate, updated coverage data on a version of the software, P <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i+1</sub> , without rerunning any test cases that do not execute the changes from the previous version of the software, Pi, to P <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i+1</sub> -Users of our technique can avoid the expense of rerunning the entire test suite on P <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i+1</sub> or the inaccuracy produced by previous approaches that estimate coverage data for P <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i+1</sub> or reuse outdated coverage data from Pi. This paper also presents a tool, RECOVER, that implements our technique, along with a set of empirical studies. The studies show the inaccuracies that can exist when an application—regression-test selection—uses estimated and outdated coverage data. The studies also show that the overhead incurred by our technique is negligible.