Detecting concurrency bugs in higher-order programs through abstract interpretation

Manually detecting bugs in concurrent programs is hard due to the myriad of thread interleavings that needs to be accounted for. Higher-order programming features only exacerbate this difficulty. The need for tool support therefore increases as these features become more widespread. We investigate the P(CEK*)S abstract machine as the foundation for tool support for detecting concurrency bugs. This abstract interpreter analyzes multi-threaded, higher-order programs with shared-store concurrency and a compare-and-swap synchronization primitive. In this paper, we evaluate two different approaches to reduce the size of the state space explored by the abstract interpreter. First, we integrate abstract garbage collection into the abstract interpreter, and we observe that it does not reduce the state space as expected. We then evaluate the impact of adding first-class support for locks on the machine's client analyses. To this end, we compare a cas-based and a lock-based formulation of race condition and deadlock detection analyses. We show that adding first-class support for locks not only significantly reduces the number of abstract program states that need to be explored, but also simplifies formulating the client analyses.