Advancing data race investigation and classification through visualization

Data races in multi-threaded programs are a common source of serious software failures. Their undefined behavior may lead to intermittent failures with unforeseeable, and in embedded systems, even life-threatening consequences. To mitigate these risks, various detection tools have been created to help identify potential data races. However, these tools produce thousands of data race warnings, often in text-based format, which makes the manual assessment process slow and error-prone. Through visualization, we aim to speed up the data race assessment process by reducing the amount of information to be investigated, and to provide a versatile interface that quality assurance engineers can use to investigate data race warnings. The ultimate goal of our integrated software suite, called RaceView, is to improve the usability of the data race information to such an extent that the elimination of data races can be incorporated into the regular software development process.