Building a Rust-style static analyzer for C++ with AI

4 months ago

The author, with 15 years of C++ experience, faced persistent memory issues like segmentation faults and memory leaks.
Rust's memory safety features were appealing, but rewriting existing C++ codebases in Rust was impractical.
Attempts to bring Rust-like memory safety to C++ via macros were deemed impossible by Google engineers.
Circle C++ offered a Rust-like borrow checker for C++ but was rejected by the C++ committee and is closed-source.
The author decided to create a static analyzer for C++ to enforce memory safety without modifying the language.
Leveraging AI coding assistants (Claude), the author developed a prototype analyzer iteratively.
The analyzer uses comment-based annotations (@safe, @unsafe) to mark code sections without altering C++ syntax.
It enforces Rust-like borrow checking, const correctness, and provides C++ equivalents of Rust types (Box, Arc, Option).
External annotations allow safe usage of STL and third-party libraries without modifying their code.
The tool integrates with build systems (CMake) and can detect violations like multiple mutable borrows or use-after-move.
The author reflects on AI's rapid evolution, noting its engineering prowess surpasses many PhD students and even their own capabilities.
The project, Rusty C++, is now stable and usable, offering peace of mind for C++ developers battling memory issues.

Hasty Briefsbeta