Using Static Analysis to Determine Where to Focus Dynamic Testing Effort

摘要

We perform static analysis and develop a negative binomial regression model to predict which files in a large software system are most likely to contain the largest numbers of faults that manifest as failures in the next release, using information from all previous releases. This is then used to guide the dynamic testing process for software systems by suggesting that files identified as being likely to contain the largest numbers of faults be subjected to particular scrutiny during dynamic testing. In previous studies of a large inventory tracking system, we identified characteristics of the files containing the largest numbers of faults and those with the highest fault densities. In those studies, we observed that faults were highly concentrated in a relatively small percentage of the files, and that for every release, new files and old files that had been changed during the previous release generally had substantially higher average fault densities than old files that had not been changed. Other characteristics were observed to play a less central role. We now investigate additional potentially-important characteristics and use them, along with the previously-identified characteristics as the basis for the regression model of the current study. We found that the top 20% of files predicted by the statistical model contain between 71% and 85% of the observed faults found during dynamic testing of the twelve releases of the system that were available.