Need LDM decisions for overload resolution in the face of 'out var'

[gafter]

We need the LDM to decide how 'out var' should interact with overload resolution. Currently, the presence of an out var argument prevents that parameter position from contributing to betterness. For example, the following program is ambiguous in the current prototype.

I would recommend saying that there is an implicit "conversion from expression" from the 'out var' to any (out) argument type. Then, the usual tie-breakers will cause the first of these methods to be selected.

class Program
{
    static void Main(string[] args)
    {
        M(out var t); // error: ambiguous?
    }
    static void M(out string s)
    {
        s = null;
    }
    static void M(out object o)
    {
        o = null;
    }
}

[gafter]

/cc @AlekseyTs FYI

[gafter]

The draft spec for "out var" says

For the purposes of overload resolution (see sections 7.5.3.2 Better function member and 7.5.3.3 Better conversion from expression), neither conversion is considered better when corresponding argument is an implicitly-typed out variable declaration.

This appears to be intended to override the parts of the current specification that compare only the method parameters, and ignore the argument and its conversions, such as

• Otherwise, if MP has more specific parameter types than MQ, then MP is better than MQ.

[gafter]

The specification in https://github.com/dotnet/roslyn/blob/master/docs/features/outvar.md has been updated in 2891628 to align with the intent of the LDM and the shape of the implementation.

[gafter]

The draft has been updated in #13999. I checked with @MadsTorgersen and the LDM does not intend to produce any further specification text for quite some time. He is OK with a spec that aligns with the implementation, the shape of the current specification, and the intent of the LDM as previously expressed. So there is nothing more to do here until it is time to integrate this into the full specification.