Race conditions are hard to detect in frameworkbased applications. Frameworks often improve performance by providing threading, but this threading is usually hidden from application programmers. Therefore, it is easy for application programmers to accidentally create data races.
Data races can be detected tracing the flow of execution, but tracing tends to produce too much data. However, the structure of frameworks can be used to control the amount of data collected and that makes tracing practical.
We have developed a tracing and analysis tool that allows application programmers to explore different configurations of an application and find probable data races. This enables them to have mutithreaded frameworks safely without having to learn the details of the framework design.
Digg
Furl
Spurl
Blink
Simpy
Google
Del.icio.us
Y!MyWeb