On Thu 05 May 2016 16:46, Alex Griffin <a...@ajgrf.com> writes: > On Thu, May 5, 2016, at 08:35 AM, Ludovic Courtès wrote: >> >> Yeah it’s OK to do it in the order: write the package bootstrapped from >> the binary blob, and then work on bootstrapping it from OCaml. > > I do not think that bootstrapping Rust from OCaml is a practical goal. > When you guys have talked about bootstrapping gcc from an older version, > I don't think it has involved even a dozen gcc versions, let alone 319! > It's a huge amount of work, to the point where it may even be easier to > write a new Rust compiler from scratch. Even if it succeeded, it would > take over a full week to build the current Rust release, meanwhile OCaml > requires a blob anyway. > > I'd love to be proven wrong, but in any case I don't think it's > reasonable to expect Jelle Licht to take on this project just because he > assumed the mantle of rustc packager.
I agree with Alex FWIW. Also FWIW, I think in the future a sane bootstrap is probably more likely from a MIR interpreter (something like this but implemented from another language: https://github.com/tsion/miri/tree/master/src). But given that even OCaml includes a binary blob (!), I don't think that we can practically require bootstrapped language implementations to bootstrap all the way. Andy
