Unused variables suggest `_prefix` for patterns with `ref` bindings when `_` wildcard would suffice

[estebank]

error: unused variable: `e`
   --> src/librustc_resolve/error_reporting.rs:234:41
    |
234 |             if let PathSource::Expr(ref e) = source {
    |                                         ^ help: consider prefixing with an underscore: `_e`
    |
    = note: `-D unused-variables` implied by `-D warnings`

The correct suggestion is either to remove the ref e and supply _e or to change the code to ref e: _.

[ehuss]

I think this may be the same as #54180.

EDIT: Or maybe not, ignore me.

[estebank]

It's very similar. It'll probably be fixed by the same code.

[petrochenkov]

remove the ref e and supply _e

This may cause a "can't move out from source" error later.

... or to change the code to ref e: _

It's longer and noisier, why is this better than ref _e?
The existing diagnostic looks like the best alternative to me.

[pnkfelix]

... or to change the code to ref e: _

It's longer and noisier, why is this better than ref _e?
The existing diagnostic looks like the best alternative to me.

Longer, noisier, and invalid: playground

The only thing I can think of that is actually valid code that looks like that in this context is Variant { 0: ref _e } (playground)

but like @petrochenkov , I do not see why one would suggest that over Variant(ref _e)

[pnkfelix]

Or maybe the intent is to suggest Variant(_) rather than Variant(ref e) ?

[pnkfelix]

In any case, if we are indeed talking about tuple-structs, I don't see how the given suggestion as implemented is "incorrect." It may not be ideal (if I'm right that Variant(_) should always be a valid option here), but striving for the ideal is a feature request, not a bug fix

[estebank]

This is what I get for creating tickets sleep deprived. I misread the error as having the problem in #54180. We can close this one.