For const variable declaration, if type is left dangling, also suggest type

[Rageking8]

Given the following code:

fn main() {
    const b: = 123;
}

The current output is:

   Compiling playground v0.0.1 (/playground)
error: expected type, found `=`
 --> src/main.rs:2:14
  |
2 |     const b: = 123;
  |              ^ expected type

error: could not compile `playground` due to previous error

Since the compiler suggests the type needed for the following:

fn main() {
    const b = 123;
}

   Compiling playground v0.0.1 (/playground)
error: missing type for `const` item
 --> src/main.rs:2:11
  |
2 |     const b = 123;
  |           ^ help: provide a type for the constant: `b: i32`

error: could not compile `playground` due to previous error

The output ideally should also suggest the type if its left dangling.

[compiler-errors]

cc @WaffleLapkin who knows how to do this since he worked on that typeof PR i think

[WaffleLapkin]

@rustbot claim

[WaffleLapkin]

The output ideally should also suggest the type if its left dangling.

@Rageking8 what you've meant by a type that's left dangling?

[Rageking8]

Basically just that the colon is included but no type is specified.
Just like in the first example (where the type is left empty, with a colon), it should also suggest the type to include to make the program compile. (Like the second example which suggests the type to include)

[chenyukang]

This is because if we have colon, the parser will try to parse a type, then error will trigger.
If we don't have colon, parser will try to insert a place hold:

const a = 123
=>
const a: _ = 123

Then type checker will enrich the type messages.

https://github.com/rust-lang/rust/blob/master/compiler/rustc_parse/src/parser/item.rs#L1183

I'm thinking on the solution.

[chenyukang]

@WaffleLapkin
I have a branch in local could fix this issue.
But it also break another testcase with wrong syntax, but I think it's OK since this testcase is far from normal syntax.
Do you mind I send a PR for this?

[WaffleLapkin]

@chenyukang oh, I've had a fix too, oops 😅

I've opened a PR already: #100168, but feel free to open a PR as well, if you think your approach might be better.

[chenyukang]

@chenyukang oh, I've had a fix too, oops 😅

I've opened a PR already: #100168, but feel free to open a PR as well, if you think your approach might be better.

My fix is shorter, but seems a little bit hacky 😁
FYI: chenyukang@3440429