Efficient use of code coverage in large-scale software development

Due to the increasing growth in complexity and size of software being developed, and the nature of large-scale software development, performing code coverage analysis is not an easy task despite the benefits to quality improvement. Our goal is to find an efficient way to perform code coverage analysis for large industrial software systems using the results from defect distribution of previous releases and the release under current development. Empirical verification of several software metrics in large-scale software development were investigated to reduce efforts required for coverage analysis and to identify error-prone modules. The metrics examined were: • coarse coverage analysis versus detailed coverage analysis • block versus decision coverage • Pareto-like defect distribution by releases, and components • cyclomatic complexity versus defect and module size The results suggested that detailed coverage analysis of all modules is not desirable and it is not cost effective. Therefore, we present a systematic approach of coverage analysis based on the findings. It will minimize the amount of effort needed for coverage analysis while maximizing the possibility of catching bugs.