Trait bound requiring the same trait with different lifetimes results in a compile error

[theemathas]

I tried this code:

trait SomeTrait<'a> {}

fn foo<'a, 'b, T: SomeTrait<'a> + SomeTrait<'b>>() {}

I expected this code to compile, but instead I got this error:

   Compiling playground v0.0.1 (/playground)
error[E0283]: type annotations needed: cannot satisfy `T: SomeTrait<'a>`
 --> src/lib.rs:3:19
  |
3 | fn foo<'a, 'b, T: SomeTrait<'a> + SomeTrait<'b>>() {}
  |                   ^^^^^^^^^^^^^
  |
note: multiple `impl`s or `where` clauses satisfying `T: SomeTrait<'a>` found
 --> src/lib.rs:3:19
  |
3 | fn foo<'a, 'b, T: SomeTrait<'a> + SomeTrait<'b>>() {}
  |                   ^^^^^^^^^^^^^   ^^^^^^^^^^^^^

For more information about this error, try `rustc --explain E0283`.
error: could not compile `playground` (lib) due to 1 previous error

Meta

Reproduces on the playground on "Stable version: 1.80.1" and "Nightly version: 1.83.0-nightly (2024-09-03 d6c8169)"

This compiles in rust 1.3.0. ICEs in rust 1.4.0-1.6.0. Fails to compile in rust 1.7.0.

[workingjubilee]

Oh.

This is #30389

[theemathas]

@workingjubilee How does this relate to RFC 1214? The RFC seems to imply that errors caused by the changes could be fixed by adding more bounds somewhere. And I can't figure out how to add such a bound for this case.

[compiler-errors]

If you want to fix this, you need to write T: for<'r> SomeTrait<'r>. This is not fixable under the way the type system works today.

[workingjubilee]

@theemathas The RFC is incomplete, I refer to this comment:

                // Typically, this ambiguity should only happen if
                // there are unresolved type inference variables
                // (otherwise it would suggest a coherence
                // failure). But given #21974 that is not necessarily
                // the case -- we can have multiple where clauses that
                // are only distinguished by a region, which results
                // in an ambiguity even when all types are fully
                // known, since we don't dispatch based on region
                // relationships.

[workingjubilee]

alternatively this could also be considered a duplicate of #21974