Extensions: proposal to use type inference during lookup #9126
Conversation
Co-authored-by: Jan Jones <[email protected]>
|
|
||
| My proposal is that the implementation continue to diverge from the spec: instead of doing 2-phase lookup | ||
| (as described in the section above, where we find compatible substituted extension containers, then find the candidate members in those) | ||
| we could do a 1-phase lookup. We would only do this in invocation scenarios (ie. when arguments are provided). |
There was a problem hiding this comment.
At this point should we just update the spec with the compiler behavior? If the behavior is so ingrained we can't change it with classic extension methods and will continue it for extension members, instead of it being a spec deviation maybe we change the spec to reflect the implementation.
There was a problem hiding this comment.
(otherwise we should document it in https://github.com/dotnet/roslyn/blob/main/docs/compilers/CSharp/Deviations%20from%20Standard.md ?)
There was a problem hiding this comment.
We'll discuss this topic in LDM tomorrow.
FWIW, I have an issue tracking the discrepancy between spec and compiler here: dotnet/csharpstandard#1203 (we'll convert to roslyn issue if we decide to adjust implementation or document deviation instead of updating spec).
Relates to test plan dotnet/roslyn#76130