Darko Marinov

Testing is the most common approach in practice to check software. Regression testing checks software changes. A key challenge for regression tests is to detect software bugs and fail when a change introduces a bug, signaling to the developer to fix it. However, an emerging challenge is that the tests must also *not* fail when there is *no* bug in the change. Unfortunately, some tests, called flaky, can non-deterministically pass or fail on the same software, even when it has no bug. Such flaky tests give false alarms to developers about the existence of bugs. A developer may end up wasting time trying to fix bugs not relevant to the recent changes the developer made. I will present some work done by my group and our collaborators to alleviate the problem of flaky tests.