Skip to Main Content
The past decade has seen exciting changes in how we develop and test software. Researchers have invented numerous techniques and criteria that are now mature enough to be ready for industrial use. During this time, the need for reliable software has grown enormously. The user base is expanding, technological advances put software into more mission-critical locations, software continues to grow in complexity, and secure software must be correct software. However, many of the strongest testing ideas invented by researchers have not yet been adopted by industry. In particular, one of the hardest problems in software testing is automatic generation of test inputs; a problem with many sophisticated solutions from the research community but for which industry tools only have primitive solutions. This talk will discuss automatic test data generation in the context of the model-driven test design process, then discuss mismatches between testing in industry and techniques from the research community. The talk will discuss why industry needs to improve testing, explore some of the difficulties in transitioning testing research results to industrial use, and close with a description of practical, usable engineering tools that can incorporate the best automatic test data generation ideas in pragmatic ways.