If the first type parameter gets inferred, that's still not handled correctly;
it'll require some more refactoring: E.g. if we have `Thing<T, F=fn() -> T>` and
then instantiate `Thing<_>`, that gets turned into `Thing<_, fn() -> _>` before
the `_` is instantiated into a type variable -- so afterwards, we have two type
variables without any connection to each other.
When referring to an associated type of a super trait, we used the substs of the
subtrait. That led to the #4931 crash if the subtrait had less parameters, but
it could also lead to other incorrectness if just the order was different.
Fixes#4931.
This is working, but I'm not that happy with how the lowering works. We might
need an additional representation between `TypeRef` and `Ty` where names are
resolved and `impl Trait` bounds are separated out, but things like inference
variables don't exist and `impl Trait` is always represented the same
way.
Also note that this doesn't implement correct handling of RPIT *inside* the
function (which involves turning the `impl Trait`s into variables and creating
obligations for them). That intermediate representation might help there as
well.
So e.g. if we have `fn foo<T: SomeTrait<u32>>() -> T::Item`, we want to lower
that to `<T as SomeTrait<u32>>::Item` and not `<T as SomeTrait<_>>::Item`.
Basically, if we had something like `dyn Trait<T>` (where `T` is a type
parameter) in an impl we lowered that to `dyn Trait<^0.0>`, when it should be
`dyn Trait<^1.0>` because the `dyn` introduces a new binder. With one type
parameter, that's just wrong, with two, it'll lead to crashes.
3964: Nicer Chalk debug logs r=matklad a=flodiebold
I'm looking at a lot of Chalk debug logs at the moment, so here's a few changes to make them slightly nicer...
3965: Implement inline associated type bounds r=matklad a=flodiebold
Like `Iterator<Item: SomeTrait>`.
This is an unstable feature, but it's used in the standard library e.g. in the definition of Flatten, so we can't get away with not implementing it :)
(This is cherry-picked from my recursive solver branch, where it works better, but I did manage to write a test that works with the current Chalk solver as well...)
3967: Handle `Self::Type` in trait definitions when referring to own associated type r=matklad a=flodiebold
It was implemented for other generic parameters for the trait, but not for `Self`.
(Last one off my recursive solver branch 😄 )
Co-authored-by: Florian Diebold <flodiebold@gmail.com>
Like `Iterator<Item: SomeTrait>`.
This is an unstable feature, but it's used in the standard library e.g. in the
definition of Flatten, so we can't get away with not implementing it :)
The big change here is counting binders, not
variables (https://github.com/rust-lang/chalk/pull/360). We have to adapt to the
same scheme for our `Ty::Bound`. It's mostly fine though, even makes some things
more clear.
It improves compile time in `--release` mode quite a bit, it doesn't
really slow things down and, conceptually, it seems closer to what we
want the physical architecture to look like (we don't want to
monomorphise EVERYTHING in a single leaf crate).
To do this we need to carry around the original resolution a bit, because `Self`
gets resolved to the actual type immediately, but you're not allowed to write
the equivalent type in a projection. (I tried just comparing the projection base
type with the impl self type, but that seemed too dirty.) This is basically how
rustc does it as well.
Fixes#3249.
The `-` turned into a `+` during a refactoring.
The original issue was caused by `Read` resolving wrongly to a trait without
type parameters instead of a struct with one parameter; this only fixes the
crash, not the wrong resolution.
