Swift at Apple: Migrating the TrueType hinting interpreter

Original Article ↗ Hacker News Discussion ↗

Apple’s rewrite of the TrueType font hinting interpreter in Swift is being seen as a flagship example of its broader push to replace C/C++ with memory-safe languages across the OS stack, including kernels and the Secure Enclave. Commenters weigh the trade-offs between Swift and Rust for systems programming, touching on ABI stability, ergonomics, and interoperability with Objective‑C and C++, while also noting the heavy reliance on exhaustive testing and fuzzing to validate such low-level changes. The work has revived debates over macOS font rendering on low‑DPI displays, licensing choices like MIT vs Apache 2.0, and the growing but carefully supervised use of LLMs in production-grade systems code.

Swift adoption and “RIS” (Rewrite In Swift)

  • Commenters note Apple is rewriting many OS components in Swift across stacks (kernels on some chips, Secure Enclave, now TrueType hinting).
  • Some wish WWDC had offered deeper technical sessions on these rewrites instead of focusing on higher-level frameworks.
  • There is curiosity and uncertainty about how much of WebKit/Safari will actually move from C++ to Swift, given recent features still landing in C++.

Swift vs. Rust and language design tradeoffs

  • Debate over an alternate history where Apple picked Rust instead of Swift.
  • Points in favor of Swift: easier ergonomics, Objective‑C/C/C++ interop, stable ABI, runtime model that fits Apple’s dynamic frameworks, and more approachable ownership options (RC + copy‑on‑write vs. strict borrow checking).
  • Rust is praised for safety and ecosystem, but seen as more complex, with no standard ABI and more friction (.clone(), back references, etc.).
  • Both languages are viewed as mutually influential and converging on similar ownership ideas; Swift lifetimes/borrow-like features are still rough for some users, with reports of compiler crashes.

Licensing and open-source posture

  • The TrueType interpreter is MIT-licensed, which people find notable given Apple’s usual preference for Apache 2.
  • Explanations: patents on TrueType are likely expired, no plans for outside contributions, and MIT’s simplicity may encourage broader reuse.

AI/LLMs in development

  • Some see “LLM smells” in the code and worry about overreliance on assistants.
  • Others state that every line and even generated assembly were heavily reviewed and tested.
  • Apple is said to use external code assistants heavily; some wish they would dogfood their own Xcode agents more.

Font hinting, DPI, and macOS rendering

  • Strong disagreement over macOS’s largely unhinted text on low‑DPI displays.
  • One side argues macOS text is blurry on 1080p and that Apple ignores common resolutions; hinting would improve legibility.
  • The other side prefers faithful letter shapes over crispness, notes macOS has long done this, and argues HiDPI largely makes the issue moot.
  • Some fonts still require hinting; Apple rewrote the hinter for security rather than fully deprecating it.

Testing, fuzzing, and quality engineering

  • Commenters highlight that the project wrote ~4× more test code than interpreter code as a sign of serious quality goals.
  • Use of fuzzers to reduce a 10M‑PDF corpus to ~4,200 for coverage is discussed; the underlying set‑cover optimization is noted as non‑trivial.
  • Some feel manual rewrites of constructs like map/filter for performance signal compiler optimization gaps.

Security and hiring

  • Apple teams are actively hiring for Swift-based systems and kernel work focused on memory safety.
  • A shared resource on low-level software security is recommended for learning to reason about vulnerabilities.