Do not ICE on malformed suggestion spans

[estebank]

Under the assumption that suggestions are by their very nature always "best effort", it is ok if we don't display them instead of having an ICE. The underlying issue of the malformed span being created is left unaddressed.

Fix #67567.

r? @petrochenkov

[estebank]

The original report produces, the following, where we see we're already emitting an invalid span in the primary diagnostic window, but handling it gracefully there:

error: expected identifier, found `)`
  --> src/ice.rs:24:9
   |
24 |         }
   |         ^ expected identifier

error: mismatched closing delimiter: `}`
  --> src/ice.rs:24:9
   |
22 |         } else {
   |                - closing delimiter possibly meant for this
23 |             return Err(m1::E1::
   |                       - unclosed delimiter
24 |         }
   |         ^ mismatched closing delimiter

error[E0308]: mismatched types
  --> src/main.rs:1:1
   |
1  | / mod ice;
2  | | fn main() {
3  | | }
...  |
   |
   = note: expected enum `std::result::Result<ice::m1::S1, ()>`
            found struct `ice::m1::S1`

Fixing the syntax error in the original report produces reasonable spans:

error[E0308]: mismatched types
  --> src/ice.rs:19:20
   |
19 |               return m1::S1 {
   |  ____________________^
20 | |                 v1: format!(""),
21 | |             }
   | |_____________^ expected enum `std::result::Result`, found struct `ice::m1::S1`
   |
   = note: expected enum `std::result::Result<ice::m1::S1, ()>`
            found struct `ice::m1::S1`
help: try using a variant of the expected enum
   |
19 |             return Ok(m1::S1 {
20 |                 v1: format!(""),
21 |             })
   |

The underlying issue is that we're taking a span from main.rs and combining it (without checking) with an end point from ice.rs.

[bors]

☔ The latest upstream changes (presumably #68272) made this pull request unmergeable. Please resolve the merge conflicts.

[petrochenkov]

@bors r+

[bors]

📌 Commit 03240e1 has been approved by petrochenkov

[pnkfelix]

discussed at T-compiler meeting. approved for beta-backport.

[Mark-Simulacrum]

@estebank This PR does not backport to beta cleanly, and the set of merge conflicts. Can you prepare a commit that backports this onto origin/beta?

[Mark-Simulacrum]

Never mind. I forgot about the rustfmt changes on master, applying those to beta allowed this to backport cleanly atop them.