`#[test]` interacts strangely with attribute macros that discard the function

I'm not sure if this is a bug or not.

dep/src/lib.rs:

use proc_macro::TokenStream;

#[proc_macro_attribute]
pub fn discard(_: TokenStream, _: TokenStream) -> TokenStream {
    TokenStream::new()
}

src/lib.rs:

#![allow(dead_code)]

fn lol() {
    panic!("why are you calling me");
}

#[test]
#[dep::discard]
fn lol() {}

Running cargo test on this gives the following output:

    Finished `test` profile [unoptimized + debuginfo] target(s) in 0.00s
     Running unittests src/lib.rs (target/debug/deps/foo-ebbeb6822a5b805f)

running 1 test
test lol ... FAILED

failures:

---- lol stdout ----

thread 'lol' (37963) panicked at src/lib.rs:4:5:
why are you calling me
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace


failures:
    lol

test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s

error: test failed, to rerun pass `--lib`

It seems that a#[test] attribute will attempt to call a function with the same name as the annotated function, whether it's the same function or not. It will even try to call a const with the same name, if one is available. This seems rather strange. Not sure what the correct behavior is though.

Meta

rustc --version --verbose:

rustc 1.93.0-nightly (292be5c7c 2025-10-29)
binary: rustc
commit-hash: 292be5c7c05138d753bbd4b30db7a3f1a5c914f7
commit-date: 2025-10-29
host: aarch64-apple-darwin
release: 1.93.0-nightly
LLVM version: 21.1.3

Reproduce the report with the examples in dep/src/lib.rs and src/lib.rs, starting with cargo test and the shown nightly toolchain. Trace how #[test] and dep::discard interact when the function disappears, including the same-name const case mentioned in the report. Done should establish the intended behavior and provide coverage for the agreed handling.