Reject integer suffix when tuple indexing

[estebank]

Fix #59418.

r? @varkor

[estebank]

Fixed indenting in the block, you can ignore the whitespace changes with https://github.com/rust-lang/rust/pull/59421/files?w=1

[varkor]

The message could probably be rephrased to sound less awkward, but r=me either way.

[varkor]

Fixed indenting in the block

I had done exactly the same thing 😄

[estebank]

@bors r=varkor

[bors]

📌 Commit 6ad77b0 has been approved by varkor

[petrochenkov]

I was going to blame beginner contributors and diagnostic improvements for this regression, but it turned out it was me - 44acea4.

[petrochenkov]

@estebank
Could you also add test cases for suffixed literals in fields in struct expressions and struct patterns.

let s = S { 0suffix: 0 };
match s {
    S { 0suffix: _ } => {}
}

[Centril]

@petrochenkov @estebank This is already r+ed, should we add the extra test cases in follow up PRs? (or do you want to r-?)

[estebank]

@bors r-

[estebank]

Added the requested changes. r? @petrochenkov, can you take a look to give final approval?

For the case you asked the output is

error: expected identifier, found `0suffix`
  --> src/main.rs:12:17
   |
12 |     let s = X { 0suffix: 0, 1: 1, 2: 2 };
   |             -   ^^^^^^^ expected identifier
   |             |
   |             while parsing this struct

error: expected identifier, found `0suffix`
  --> src/main.rs:15:13
   |
15 |         X { 0suffix: _, .. } => {}
   |             ^^^^^^^ expected identifier

error[E0063]: missing field `0` in initializer of `X`
  --> src/main.rs:12:13
   |
12 |     let s = X { 0suffix: 0, 1: 1, 2: 2 };
   |             ^ missing `0`

error[E0027]: pattern does not mention fields `0`, `1`, `2`
  --> src/main.rs:15:9
   |
15 |         X { 0suffix: _, .. } => {}
   |         ^^^^^^^^^^^^^^^^^^^^ missing fields `0`, `1`, `2`
   |
   = note: trying to match a tuple variant with a struct variant pattern

Which is why I think it shouldn't be added to the test, as it is not handled by this PR.

[petrochenkov]

@estebank

Which is why I think it shouldn't be added to the test, as it is not handled by this PR.

I think what this PR is doing (accept suffixed literal, report a non-fatal error, continue) is a better alternative than "expected identifier, found 0suffix", so we can use it for struct expressions/patterns as well and unify the diagnostics.

[estebank]

@petrochenkov

I see. Fixed.

[petrochenkov]

Thanks!
@bors r+

[bors]

📌 Commit 8d1cc72 has been approved by petrochenkov

[kpp]

How about X { 0usize: 0, 1: 1, 2: 2 };?

[Centril]

(don't r- the PR in either case since it's in a rollup now... if changes need to be made, do it in a follow up PR)