regression: mismatched types (invalid pointer cast for trait objects)

[BoxyUwU]

• https://crater-reports.s3.amazonaws.com/beta-1.81-2/beta-2024-07-26/reg/minfac-0.0.1/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.81-2/beta-2024-07-26/reg/saleae-0.1.0/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.81-2/beta-2024-07-26/reg/vsmtp-common-2.2.1/log.txt
• https://crater-reports.s3.amazonaws.com/beta-1.81-2/beta-2024-07-26/reg/vsmtp-protocol-2.2.1/log.txt

[INFO] [stdout] error[E0308]: mismatched types
[INFO] [stdout]   --> src/untyped.rs:26:11
[INFO] [stdout]    |
[INFO] [stdout] 26 |         &*(self.pointer as *const dyn Fn(&ServiceProvider) -> T)
[INFO] [stdout]    |           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected `()`, found `(&ServiceProvider,)`
[INFO] [stdout]    |
[INFO] [stdout]    = note: expected trait object `dyn Fn()`
[INFO] [stdout]               found trait object `dyn for<'a> Fn(&'a ServiceProvider) -> T`

#120248 (comment) I believe is accepted breakage? cc @WaffleLapkin

[compiler-errors]

Yeah, this is known breakage I think. Or if it wasn't uncovered by the crater, I totally think we should reject it.

[WaffleLapkin]

I agree, all of this looks like code we intended to break.

The errors are unfortunately opaque though, I imagine users would have hard time figuring them out :(

[VorpalBlade]

@WaffleLapkin shouldn't this be turned into a diagnostic issue rather than closed if the error is subpar (which it looks like it is to me as well)?

[WaffleLapkin]

@VorpalBlade diagnostic issue should be opened, but I have no spoons for that. Turning issues into issues for other topics is generally confusing, opening new issues is better.

[apiraino]

@VorpalBlade if you have capacity, please feel free to open such issue about diagnostic (backlinking this one). Thanks!

[VorpalBlade]

I have the capacity, but I don't have any idea what the diagnostic should be. I don't understand the current error message.

[riking]

The error is that someone's trying to smuggle Fns of several different argument sets through a single variable of type *const dyn Fn() (no arguments) without a transmute.

[WaffleLapkin]

@apiraino @VorpalBlade I've opened #130030 to track the diagnostic issue.