Traits in type position should be a fatal error in ≥2018 editions

[kpreid]

Code

struct Foo {
    a: Fn(),
    b: Fn(),
}

Current output

error[E0277]: the size for values of type `(dyn Fn() + 'static)` cannot be known at compilation time
 --> src/lib.rs:2:8
  |
2 |     a: Fn(),
  |        ^^^^ doesn't have a size known at compile-time
  |
  = help: the trait `Sized` is not implemented for `(dyn Fn() + 'static)`
  = note: only the last field of a struct may have a dynamically sized type
  = help: change the field's type to have a statically known size
help: borrowed types always have a statically known size
  |
2 |     a: &Fn(),
  |        +
help: the `Box` type always has a statically known size and allocates its contents in the heap
  |
2 |     a: Box<Fn()>,
  |        ++++    +

error[E0782]: trait objects must include the `dyn` keyword
 --> src/lib.rs:2:8
  |
2 |     a: Fn(),
  |        ^^^^
  |
help: add `dyn` keyword before this trait
  |
2 |     a: dyn Fn(),
  |        +++

error[E0782]: trait objects must include the `dyn` keyword
 --> src/lib.rs:3:8
  |
3 |     b: Fn(),
  |        ^^^^
  |
help: add `dyn` keyword before this trait
  |
3 |     b: dyn Fn(),
  |        +++

Desired output

There are two fixes that I think would be wise here.

1. Most importantly, the E0277 should not occur; the compiler should not
   attempt to continue interpreting the program as if it were in edition 2015. Or, at a minimum, the E0782 should appear first.

2. E0782 could be better expressed, in the modern perspective, as “expected type, found trait”. Don't bring up trait objects at all until the author writes dyn.

Rationale and extra context

Today, when a newcomer writes a trait in a type position, it most often means they're confused about what the difference is, not that they wanted a trait object and need to be taught the new syntax. The whole point of adding dyn to Rust's syntax was to remove that easy confusion. Therefore, the compiler should stop making the assumption whatsoever that the author of the code meant a trait object.

This will have the further advantage that newcomers will not be confronted with errors about Sized when they probably didn't even mean to use any DST.

Rust Version

rustc 1.81.0-nightly (ba1d7f4a0 2024-06-29)
binary: rustc
commit-hash: ba1d7f4a083e6402679105115ded645512a7aea8
commit-date: 2024-06-29
host: x86_64-apple-darwin
release: 1.81.0-nightly
LLVM version: 18.1.7

[GrigorenkoPV]

cc #127689 #127690 #127691

[kpreid]

Thanks for the cross-references. I'd like to emphasize that what I'm asking for, compared to those, is not just “make this error more concise and clear”, but “stop implicitly assuming the user definitely wanted a trait object”.

(Perhaps dyn could still be mentioned in help, but in that case, impl Trait/generics should be mentioned at the same time as they are both ways of “I want to accept an implementor of a trait here”.)

[GrigorenkoPV]

“stop implicitly assuming the user definitely wanted a trait object”.

I could not agree more. I myself have run into this issue about 3 times and 3 times out of 3 I actually meant impl Trait.

[VulnBandit]

@rustbot claim
I think this vulnerability can be addressed by me.