Should_panic can be applied to non-tests

[JonathanBrouwer]

I tried this code:

#[should_panic]
pub fn main() {}

I expected this to produce a warning or error.

Instead, it produces no warnings/errors.

Meta

rustc --version --verbose:

cargo 1.90.0-nightly (930b4f62c 2025-06-28)
release: 1.90.0-nightly
commit-hash: 930b4f62cfcd1f0eabdb30a56d91bf6844b739bf
commit-date: 2025-06-28
host: x86_64-unknown-linux-gnu
libgit2: 1.9.0 (sys:0.20.2 vendored)
libcurl: 8.12.1-DEV (sys:0.4.80+curl-8.12.1 vendored ssl:OpenSSL/3.5.0)
ssl: OpenSSL 3.5.0 8 Apr 2025
os: Arch Linux Rolling Release [64-bit]

@rustbot claim

[jdonszelmann]

lang team: should this be a warning or an error from now on? i.e. is it reasonable to assume nobody does this right now despite being allowed. This pr is blocked on the decision: #143808 (review)

[traviscross]

In my estimation, we're going to at least have appetite for making this a warning. To decide about whether it could immediately become a hard error, I'd estimate we're going to need to see a crater run.

If you want to start with the FCW (which we could later upgrade), you can nominate the PR doing that for us. If you'd like us to consider going straight to deny-by-default or a hard error, please nominate the PR with the crater results.