4 ms·
It also looks rather incomplete. Are they not planning on implementing borrowck? If chalk/polonius were ready and had a C API the roadmap would begin to make se
by unanswered 6y ago
It also looks rather incomplete. Are they not planning on implementing borrowck? If chalk/polonius were ready and had a C API the roadmap would begin to make sense.
- Nullabillity 6y agoMakes sense to me. As mrustc[0] mentions, implementing validation in a secondary compiler is much less important, because you can always just run the reference implementation as a glorified linter in the meantime. [0]: https://github.com/thepowersgang/mrustc https://github.com/thepowersgang/mrustc
- unanswered 6y agoThat's what I don't get. I looked through the available documentation briefly and didn't see any mention that this is intended to be a `mrustc`; it really seems to want to be a `rustc`. I'm not aware of other GCC frontends being less than complete compilers for their respective languages, and while I think that "rust without borrowck" is an interesting point in design space (discriminated unions, generics, macros, traits, closures), "rust without borrowck" is not rust.
- bobthebuilders 6y agoYou can plug in the borrow checker after writing the base compiler, but you if you block waiting for the borrow checker you get nothing done.