E0613 needs to be merged into E0609 now

[est31]

With RFC 1506 merged and stabilized, handling E0613 separate from E0609 is not the correct behaviour any more.

For context, you get E0613 when doing something like let _ = "hi".0;:

error[E0613]: attempted to access tuple index `0` on type `&'static str`, but the type was not a tuple or tuple struct

And you get E0609 when doing something like let _ = "hi".foo;:

error[E0609]: no field `foo` on type `&'static str`

RFC 1506 "clarified ADT kinds" (Rendered, tracking issue) allowed all structs to contain fields with integral names like 0 or 1, etc. This makes the error message for E0613 outdated, as the type may be a non tuple struct, and calling 0 a "tuple index" sounds wrong as well.

So I propose E0613 to be merged into E0609, as E0609 is now better suiting, and also because its "in the spirit" of the RFC.

The proposed new output for let _ = "hi".0; is:

error[E0609]: no field `0` on type `&'static str`

[nikomatsakis]

Makes sense to me!

[nikomatsakis]

Here are some brief mentoring instructions:

• E0613 is issued here, in librustc_typeck.
• E0609 is issued here.
• We basically need to introduce a common helper function and call it from both places. A helper is needed because you can't have more than one place in the source code that uses a single error code (in this case, E0609).

[strangeglyph]

Hi! I'd like to take this issue, if no one minds?

[Mark-Simulacrum]

Go right ahead! Let us know if you'd like additional instructions and we'll try to connect you with @nikomatsakis or someone else from the compiler team.

[GuillaumeGomez]

Or I can as well since I'm the one who added these two error codes.

[strangeglyph]

While I'm at it, E0612 looks like it could be folded into E0609 too. Should I go ahead and do that?

[est31]

While I'm at it, E0612 looks like it could be folded into E0609 too.

I can be, not sure whether it should. It seems we can actually recognize when we have a tuple struct, so we won't generate the error if it is wrongly placed. However it might be against the spirit of the RFC. @petrochenkov as author of the RFC, what do you think, should E0612 be merged into E0609?

[petrochenkov]

@est31
I'd say yes, because I'm generally for removing code and special cases in case of doubt, but people managing diagnostics often hold the opposite position.

[strangeglyph]

Should unused diagnostic codes just be removed, or commented out with a reference that the diagnostic has been folded into another?

EDIT: For now, left a comment that they have been merged into E0609