Higher kinded checks are deferred to the time of use

[hkalbasi]

In this code I will get a compiler error:

struct F where i32: Iterator;

error[[E0277]](https://doc.rust-lang.org/stable/error-index.html#E0277): `i32` is not an iterator
 --> src/main.rs:2:20
  |
2 |     struct F where i32: Iterator;
  |                    ^^^^^^^^^^^^^ `i32` is not an iterator
  |
  = help: the trait `Iterator` is not implemented for `i32`
  = note: if you want to iterate between `start` until a value `end`, use the exclusive range syntax `start..end` or the inclusive range syntax `start..=end`
  = help: see issue #48214

But in this code there is no compiler error:

struct F where for<'a> &'a i32: 'static;

But if I use F, I will get a compiler error:

struct F where for<'a> &'a i32: 'static;
let _x = F;

error: higher-ranked lifetime error
 --> src/main.rs:4:14
  |
4 |     let _x = F;
  |              ^
  |
  = note: could not prove `for<'a> &'a i32: 'static`

error: could not compile `playground` due to previous error

Which is inconsistent behavior. I think either both of them should compile, or both of them should not.

[compiler-errors]

This comment perhaps sheds some light on this inconsistency being (somewhat) intentional: #100090 (comment)

[hkalbasi]

So after feature trivial_bounds it will become a warning for both cases?

[QuineDot]

With trivial_bounds, the Iterator case would behave similar to the higher-ranked case (able to compile the definition but not to call it due to unsatisfiable bounds). And I believe the idea is both would warn about the unsatisfiable definition, yes.