Extra trait bound makes function body fail to typecheck

[jplatte]

I tried this code (playground):

pub trait Foo<'a> {}

impl Foo<'_> for () {}
impl<'a, T> Foo<'a> for Option<T> where T: Foo<'a> {}

pub fn bar<T>()
where
    T: for<'a> Foo<'a>,
{
}

fn baz<'a>()
where
    (): Foo<'a>,
{
    bar::<Option<()>>();
}

Or with references instead of Option

impl<'a, 'b, T> Foo<'a> for &'b T where T: Foo<'a> {}
bar::<&()>();

I expected to see this happen: Compile successfully, with or without the unneeded (): Foo<'a> bound.

Instead, this happened: It works when the (): Foo<'a> bound is removed, but as written it produces:

error: implementation of `Foo` is not general enough
  --> src/lib.rs:16:5
   |
1  | pub trait Foo<'a> {}
   | -------------------- trait `Foo` defined here
...
16 |     bar::<Option<()>>();
   |     ^^^^^^^^^^^^^^^^^ implementation of `Foo` is not general enough
   |
   = note: `Option<()>` must implement `Foo<'0>`, for any lifetime `'0`...
   = note: ...but `Option<()>` actually implements `Foo<'1>`, for some specific lifetime `'1`

Meta

rustc --version --verbose:

rustc 1.50.0 (cb75ad5db 2021-02-10)
binary: rustc
commit-hash: cb75ad5db02783e8b0222fee363c5f63f7e2cf5b
commit-date: 2021-02-10
host: x86_64-unknown-linux-gnu
release: 1.50.0

Also happens on Nightly.

[jplatte]

@rustbot label: +A-traits

[BoxyUwU]

theres a zulip topic in wg-traits about this here

[bsodmike]

I'm on rustc 1.59.0-beta.5 (28c8a34e1 2022-01-27) and am getting this as well,

error: implementation of `sqlx::Decode` is not general enough
   --> src/models.rs:183:5
    |
183 |     Decode,
    |     ^^^^^^ implementation of `sqlx::Decode` is not general enough
    |
    = note: `std::string::String` must implement `sqlx::Decode<'0, sqlx::Postgres>`, for some specific lifetime `'0`...
    = note: ...but it actually implements `sqlx::Decode<'r, sqlx::Postgres>`
    = note: this error originates in the derive macro `Decode` (in Nightly builds, run with -Z macro-backtrace for more info)

[grv07]

I am on 1.65 and still getting this :(

[frederikhors]

Is there a way to close this? This is very huge.