Rust to C compiler – 95.9% test pass rate, odd platforms

Original Article ↗ Hacker News Discussion ↗

C↔Rust Translation and Safety

  • Several commenters wish for C→Rust tools (e.g. DARPA’s TRACTOR, c2rust) mainly to:
    • Remove C toolchains from builds.
    • Enable LTO across translated C and Rust.
    • Use translation as a first step toward idiomatic Rust rewrites.
  • Others argue automatic C→Rust must be largely unsafe and non‑idiomatic because C omits key semantic information Rust requires.
  • A WASM→Rust example shows a path to fully “safe” Rust that is still semantically sandboxed C.

Does a Rust→C Backend Preserve Rust’s Guarantees?

  • One side: if the Rust frontend has already proven safety, the backend (LLVM, C, machine code) is just another compilation step; safety is a property of the Rust source, not the target.
  • Concerns:
    • C UB mismatches (e.g. signed overflow, alias rules) can break Rust’s semantics if not handled carefully.
    • Transitivity of guarantees depends on “lossless” translation and bug‑free compilers, which is unproven.
  • Author notes workarounds for “simple” UB, uses -fsanitize=undefined, and C escape hatches like memcpy to avoid strict aliasing.

Why Compile Rust to C?

  • Primary motivation: reach platforms where Rust/LLVM don’t exist but C compilers do (NonStop, obscure microcontrollers, some consoles, proprietary OSes).
  • Also:
    • Leverage mature C tooling (static/dynamic analyzers, Frama‑C, KLEE, CompCert).
    • Possibly integrate into ecosystems that only accept C sources.
  • Skeptics say this doesn’t improve FFI interop per se; you still call C functions either way.

Platform Coverage and Alternate Backends

  • This project is a rustc backend using its own IR, initially for .NET; C support was added because the IR mapped cleanly.
  • Interest from people maintaining alpha, hppa, m68k, sh4, etc., frustrated that other backends (e.g. gcc‑based) haven’t fully bootstrapped Rust on these.
  • Debate over Rust’s target tiers: explicit but conservative guarantees vs. C compilers’ de‑facto, less‑documented support on obscure architectures.

Language Ecosystem, Learning, and Stability

  • Extended side‑discussion:
    • Rust praised for free, unified documentation, safety model, and package manager; C/C++ defended for decades of teaching material and huge ecosystems.
    • Arguments over dependency bloat vs. the cost and correctness of rolling your own libs.
    • Rust criticized by some for complexity, build times, lack of a formal standard and multiple implementations, and weaker dynamic linking story; others counter that C/C++ also evolve and require similar care.

Project Status and Quality

  • Earlier README numbers (~80% C test pass rate) were outdated; updated figures:
    • ~96% for .NET core tests, ~95.6% for C core tests.
  • Some view the post as an in‑progress status report rather than a finished product; others praise the ambition and rapid iteration, while a few question using tools that don’t yet pass 100% of tests.