Effective static race detection for Java

Abstract

We present a novel technique for static race detection in Java programs, comprised of a series of stages that employ a combination of static analyses to successively reduce the pairs of memory accesses potentially involved in a race. We have implemented our technique and applied it to a suite of multi-threaded Java programs. Our experiments show that it is precise, scalable, and useful, reporting tens to hundreds of serious and previously unknown concurrency bugs in large, widely-used programs with few false alarms.

Static Object Race Detection

“…As a result, the running times reported for points-to analysis are approximately twice the running times of Spark. Our suite consists of benchmarks used in previous work on concurrency [15,9,14]. Column #Meth in Table 1 gives the size of the benchmarks in terms of the number of methods (user and library) reachable by Spark.…”

“…Benchmarks tsp through weblech are whole-programs, and jdbm through jtds are libraries. For the libraries, we converted the single-threaded harnesses from [9] to the multithreaded model described in Section 4.3.…”

“…We compared our analysis with Chord [9], the leading static race detector. Chord's data race report includes a field-based view and an object-based view of data races.…”

“…if root ∈ minBi j [7] if reentrant (i j) [8] St += (j.n, L(i.m j.n)) [9] S t += (j.n, L(i.m j.n)) [10] else [11] St += (j.n, L(i.m j.n)) [12] if j.n not visited then W += j.n…”

“…Our analysis presents relatively low false-positive rate and runs in less than 1 minute on all but one benchmark. On most benchmarks our analysis performs comparably to Chord [9], the leading static data race detector. On several benchmarks our analysis outperforms Chord, in some cases significantly.…”