Integer type inference confusion

[alexcrichton]

A program like:

fn foo(_: u8) {}

fn main() {
  let a = 1;
  foo(a.to_u8());
  foo(a);
}

yields the error

$ rustc foo.rs
foo.rs:5:6: 5:16 error: multiple applicable methods in scope
foo.rs:5   foo(a.to_u8());
               ^~~~~~~~~~
foo.rs:5:6: 5:16 note: candidate #1 is `core::i32::__extensions__::to_u8`
foo.rs:5   foo(a.to_u8());
               ^~~~~~~~~~
foo.rs:5:6: 5:16 note: candidate #2 is `core::i8::__extensions__::to_u8`
foo.rs:5   foo(a.to_u8());
               ^~~~~~~~~~
foo.rs:5:6: 5:16 note: candidate #3 is `core::i16::__extensions__::to_u8`
foo.rs:5   foo(a.to_u8());
               ^~~~~~~~~~
foo.rs:5:6: 5:16 note: candidate #4 is `core::u64::__extensions__::to_u8`
foo.rs:5   foo(a.to_u8());
               ^~~~~~~~~~
foo.rs:5:6: 5:16 note: candidate #5 is `core::u16::__extensions__::to_u8`
foo.rs:5   foo(a.to_u8());
               ^~~~~~~~~~
foo.rs:5:6: 5:16 note: candidate #6 is `core::uint::__extensions__::to_u8`
foo.rs:5   foo(a.to_u8());
               ^~~~~~~~~~
foo.rs:5:6: 5:16 note: candidate #7 is `core::int::__extensions__::to_u8`
foo.rs:5   foo(a.to_u8());
               ^~~~~~~~~~
foo.rs:5:6: 5:16 note: candidate #8 is `core::u32::__extensions__::to_u8`
foo.rs:5   foo(a.to_u8());
               ^~~~~~~~~~
foo.rs:5:6: 5:16 note: candidate #9 is `core::u8::__extensions__::to_u8`
foo.rs:5   foo(a.to_u8());
               ^~~~~~~~~~
foo.rs:5:6: 5:16 note: candidate #10 is `core::i64::__extensions__::to_u8`
foo.rs:5   foo(a.to_u8());
               ^~~~~~~~~~
foo.rs:6:6: 6:7 error: mismatched types: expected `u8` but found `i32` (expected u8 but found i32)
foo.rs:6   foo(a);
               ^
error: aborting due to 2 previous errors

I'm not really sure what's quite going on here. The first function invocation looks like it doesn't know what type a is, so it could be anything.

The second invocation I thought would force the compiler to infer that a is a u8, but it seems to have determined it's an i32 at that point which I thought would have suppressed the first error?

[catamorphism]

Reproduced with 0252c30 -- nominating for milestone 5, production-ready

[graydon]

closing INVALID, this is actually correct behavior; using method-presence to pick integer types winds up randomly causing truncation, we used to do this and stopped.