arXiv subjects

Xavier Denis

Publications and source records attributed to Xavier Denis.

3 recordsLinked to original sources

arXiv2025

Toward Practical Deductive Verification: Insights from a Qualitative Survey in Industry and Academia

Deductive verification is an effective method to ensure that a given system exposes the intended behavior. In spite of its proven usefulness and feasibility in selected projects, deductive verification is still not a mainstream technique. To pave the way to widespread use, we present a study investigating the factors enabling successful applications of deductive verification and the underlying issues preventing broader adoption. We conducted semi-structured interviews with 30 practitioners of verification from both industry and academia and systematically analyzed the collected data employing a thematic analysis approach. Beside empirically confirming familiar challenges, e.g., the high level of expertise needed for conducting formal proofs, our data reveal several underexplored obstacles, such as proof maintenance, insufficient control over automation, and usability concerns. We further use the results from our data analysis to extract enablers and barriers for deductive verification and formulate concrete recommendations for practitioners, tool builders, and researchers, including principles for usability, automation, and integration with existing workflows.

cs.SE↗

arXiv2024

Reasoning about Interior Mutability in Rust using Library-Defined Capabilities

Existing automated verification techniques for safe Rust code rely on the strong type-system properties to reason about programs, especially to deduce which memory locations do not change (i.e., are framed) across function calls. However, these type guarantees do not hold in the presence of interior mutability (e.g., when interacting with any concurrent data structure). As a consequence, existing verification techniques for safe code such as Prusti and Creusot are either unsound or fundamentally incomplete if applied to this setting. In this work, we present the first technique capable of automatically verifying safe clients of existing interiorly mutable types. At the core of our approach, we identify a novel notion of implicit capabilities: library-defined properties that cannot be expressed using Rust's types. We propose new annotations to specify these capabilities and a first-order logic encoding suitable for program verification. We have implemented our technique in a verifier called Mendel and used it to prove absence of panics in Rust programs that make use of popular standard-library types with interior mutability, including Rc, Arc, Cell, RefCell, AtomicI32, Mutex and RwLock. Our evaluation shows that these library annotations are useful for verifying usages of real-world libraries, and powerful enough to require zero client-side annotations in many of the verified programs.

cs.PL↗

arXiv2024

A Hybrid Approach to Semi-automated Rust Verification

We propose a hybrid approach to end-to-end Rust verification where the proof effort is split into powerful automated verification of safe Rust and targeted semi-automated verification of unsafe Rust. To this end, we present Gillian-Rust, a proof-of-concept semi-automated verification tool built on top of the Gillian platform that can reason about type safety and functional correctness of unsafe code. Gillian-Rust automates a rich separation logic for real-world Rust, embedding the lifetime logic of RustBelt and the parametric prophecies of RustHornBelt, and is able to verify real-world Rust standard library code with only minor annotations and with verification times orders of magnitude faster than those of comparable tools. We link Gillian-Rust with Creusot, a state-of-the-art verifier for safe Rust, by providing a systematic encoding of unsafe code specifications that Creusot can use but cannot verify, demonstrating the feasibility of our hybrid approach.

cs.PL↗