Page - 6 - in The Future of Software Quality Assurance

6 S. Amann and E. Jürgens DevelopmentVersion A Version B Existing Tests … … Missing Tests Test-Gap Analysis ImpactedTests Test-Impact Analysis Fig. 3 Change-driven testing: testing efficiently by running only the tests impacted by changes. Testing effectively by testing all changes two versionsof our system, e.g., two consecutive releases of the system beforeand after implementingaparticular feature,we proceedin twophases: 1. To make our testing more efficient, we test just the changes, excluding all other parts of the system. We automatically identify the existing tests that are relevant to thechanges, i.e., the impacted tests, througha Test-ImpactAnalysis. 2. To make our testing moreeffective, we ensure that we test all changes.To guide us towards this goal and to verify whether we achieved it, we automatically identifychanges that lack tests througha Test-GapAnalysis. Both analyses are only interested in changes that require testing. Therefore, we usestaticcodeanalyses to identifyrelevantchanges.Weconsiderachangerelevant if it modifies the behavior of the code and, thus, may contain mistakes that later cause errors. Consequently, we filter out changes that correspond to refactorings, suchaschanges todocumentationandrenamingormovingofmethodsorvariables. 3.1 Test-Impact Analysis Figure4depicts theprocessof theTest-ImpactAnalysis (TIA).Itcombinesrelevant codechanges(seeSect.2.1)with test runtimesandtest-wisecoverage(seeSect.2.3)