bivariance means that arguments are completely unconstrained

[lcnr]

struct Foo<T: Trait<Assoc = U>, U>(T);

trait Trait {
    type Assoc;
}

impl Trait for for<'a> fn(&'a ()) {
    type Assoc = String;
}

impl Trait for fn(&'static ()) {
    type Assoc = [usize; 3];
}

fn use_covariance_of_t(x: Foo<for<'a> fn(&'a ()), String>) -> Foo<fn(&'static ()), [usize; 3]> {
    x
}

[theemathas]

This code now gives the following error on version 1.76.0 stable, which seems to indicate that the behavior is a bug:

warning: conflicting implementations of trait `Trait` for type `for<'a> fn(&'a ())`
  --> src/lib.rs:11:1
   |
7  | impl Trait for for<'a> fn(&'a ()) {
   | --------------------------------- first implementation here
...
11 | impl Trait for fn(&'static ()) {
   | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ conflicting implementation for `for<'a> fn(&'a ())`
   |
   = warning: this was previously accepted by the compiler but is being phased out; it will become a hard error in a future release!
   = note: for more information, see issue #56105 <https://github.com/rust-lang/rust/issues/56105>
   = note: this behavior recently changed as a result of a bug fix; see rust-lang/rust#56105 for details
   = note: `#[warn(coherence_leak_check)]` on by default

See #56105

[lcnr]

👋 we intend to actually intend to accept this pattern long-term, so this warning will get end up getting removed without causing this code to error. We recently weakened the warning in rust-lang/rust#120716 to "the behavior may change in a future release".