Managing Post-Development Fault Removal

In this paper, we manage fault removal by classifying and prioritizing fault warnings reported by a static analysis tool. We present our findings from analyzing three cross-platform industrial code bases at Yahoo! totaling approximately 3.6+ MLOC. The tool found 1.2K potential fault warnings as follows: 52.29% true faults and 47.71% false/noise. The 52.29% correctly reported faults were prioritized based on severity. Additionally, we connected the tool classification to a standard software weakness schema, Common Weakness Enumeration (CWE) to standardized discourse. The results from creating a management system for post-development fault removal are intended to be shifted back into earlier stages of software development.