Optimization of smoke testing through data and knapsacks

This report seeks to develop a offline program that continually updates smoke testing for a large codebase in order to produce a rapidly-evolving smoke test that is completely data driven. The program, named Smoke Selector, looks to test newly implemented code by determining the code line changes on updated files. After that the Smoke Selector does two things: identifies which unit tests cover (or mostly cover) the updated lines of code and does a maximization of all the tests that will allow for the most coverage that fits under the determined time limit for the smoke test. This program fits on top of the nightly regression testing to allow a custom smoke test to be created at the beginning of the day that will test the most code on every integration as well as testing the code that is most recently changed