Type inference regression with zipped iterators

[larsluthman]

The following code compiles with Stable and Beta but not with Nightly on https://play.rust-lang.org/ as of 2015-10-24:

fn main() {
    let mut output = vec![Vec::<i32>::new(); 5];
    let input = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13];
    for chunk in input.chunks(5) {
        for (i, o) in chunk.iter().zip(output.as_mut()) {
            o.push(*i);
        }
    }
    println!("{:?}", output);
}

With Stable and Beta I get the correct output:

[[0, 5, 10], [1, 6, 11], [2, 7, 12], [3, 8, 13], [4, 9]]

With Nightly I get the error message

<anon>:6:13: 6:23 error: the type of this value must be known in this context
<anon>:6             o.push(*i);

[eefriedman]

This isn't a type inference regression; the problem is that Vec acquired a couple implementations of AsMut, and they're impossible to distinguish based on only the information "the result must implement IntoIter".

[bluss]

AsMut added in #28663. Unfortunately the smallest new trait impls cause issues like this 😞

I don't want to shift the blame, but AsMut is a conversion trait, used for generics. It shouldn't have been in the prelude, because it's not intended to be called like this IMO.

More idiomatic would be to write chunk.iter().zip(&mut output), which continues to compile.

[alexcrichton]

triage: I-nominated

[brson]

This is on beta now.

[alexcrichton]

The libs team discussed this during triage today, and the conclusion was to close this for now. This is indeed breakage with the addition of AsMut for Vec<T>, but it is covered under our API evolution policy which is to say that a local elaboration (e.g. type annotations) will fix this. If this becomes a widespread problem we will consider reverting, but for now it seems minor enough to not warrant that.