regression: #[forbid(unreachable_code)] incompatible with the Try ('?') operator

[ghost]

I tried this code:

#![forbid(unreachable_code)]

#[allow(dead_code)] // Remove warning.
fn foo() -> Option<()> {
    Some(0)?;
    None // Just so it doesn't emit error here.
}

Both on debug and release it generates an error saying that #[allow(unreachable_code)] was overridden by the #[forbid(...)] attribute, while pointing at the "try" expression.

Meta

rustc --version --verbose:

rustc 1.46.0 (04488afe3 2020-08-24)
binary: rustc
commit-hash: 04488afe34512aa4c33566eb16d8c912a3ae04f9
commit-date: 2020-08-24
host: x86_64-pc-windows-msvc
release: 1.46.0
LLVM version: 10.0

[LeSeulArtichaut]

According to Godbolt this is the case since 1.22.0 where the Try trait was implemented on Option.

[LeSeulArtichaut]

This bug can also be reproduced with a Result:

#![forbid(unreachable_code)]

#[allow(dead_code)] // Remove warning.
fn foo() -> Result<(), ()> {
    Ok(0)?;
    Err(()) // Just so it doesn't emit error here.
}

Which allows to find out that this regressed in 1.16.0.

[spastorino]

Assigning P-low as discussed as part of the Prioritization Working Group procedure and removing I-prioritize.

[Dylan-DPC]

This code now compiles fine, we just need a test added against this to close the issue

[RodBurman]

Agreed with release 1.83 the sample code compiles with no errors or warnings. Has the test been added?