Testing of a software build. Differences between software builds are tracked by scanning the binaries of a software product to automatically discover its classes. A detailed dictionary is built that captures static and dynamic information of that build, including class dependencies. A comparison may be made with another build, so that selective tests may be automatically executed on any types, and their dependencies, that have had a structural or behavioral modification since the last build. Testers may load a set of constructors for any specific types to further increase coverage of types (or classes) tested. Detailed reports may also be provided that may be used to drive future testing work and target specific areas of the code for additional testing. Code generation from intermediate code to specified targets may also be performed to aid in reproducing and fixing bugs.