Determining the adequacy of software tests, and where testing gaps might lie, is crucial for improving and maintaining the strength of test suites. Mutation analysis facilitates this by evaluating tests against generated program faults; however, no mature mutation analysis tooling exists for the safety-focused Rust systems programming language, to date. This paper introduces mutest-rs, a mature, end-to-end mutation analysis tool for Rust programs that is based on extensive static program analysis, and integrates directly with the rustc Rust compiler. Our tool overcomes the numerous challenges of generating valid Rust code mutants, and does so efficiently through a Rustspecific meta-mutant approach. Our open-source tool, mutest-rs, is available online at https://mutest.rs. A video demonstrating mutest-rs is available at https://youtu.be/8yEYAU6P63I.

Rust is a relatively young, memory safe systems programming language which is increasingly being adopted by projects requiring both performance, and safety. While automated testing is built into the language, tool support for mutation analysis is almost non-existent, having not been the subject of past research. This leaves Rust developers without a way to determine test thoroughness. To address this problem, we design a mutation analysis process for Rust that overcomes challenges related to generating viable mutations due to the strictness of the language in terms of its type system, memory restrictions, and the potential to introduce undefined behavior in unsafe code blocks. Our technique efficiently evaluates mutations simultaneously through a process we refer to as "batching" — the use of static analysis to determine mutations that are non-conflicting, and therefore are able to be evaluated together. Batching enables our technique to maximize thread usage, executing more tests in parallel, and further reducing the time required to evaluate mutations. We implemented these techniques into a tool, mutest-rs, which we empirically evaluated on a diverse set of common subject libraries and Rust programs, and found that our batching method for increasing parallelism is able to reduce the overall runtime of mutation analysis by up to 66.4%, compared to not applying batching.