`cfg_select` is not marking selected branches as expansion

[blyxyas]

Playground link

I tried this code:

fn main() {
    cfg_select! {
        true => {
            let a = 1 + 1;
        }
        false => {
            let b = 1 + 1;
        }
    }
}

I expected to see this happen: The selected branch should be marked as an expansion, so Span::from_expansion should return true.

Instead, this happened: The selected branch is pasted directly, without being marked as a macro expansion (or any expansion of any kind). This produces a bug in Clippy.

Meta

rustc --version --verbose:

rustc 1.100.0-nightly (c54751567 2026-08-22)
binary: rustc
commit-hash: c54751567b19c4ceb08b0412d83529c2568cba8b
commit-date: 2026-08-22
host: x86_64-unknown-linux-gnu
release: 1.100.0-nightly
LLVM version: 23.1.0

cc rust-lang/rust-clippy#17498 and rust-lang/rust-clippy#17508
cc @folkertdev since you were to one to implement it, maybe you can shine some insights!! (ฅ`･ω･´)っ=

[petrochenkov]

The selected branch should be marked as an expansion

The code let a = 1 + 1; does not come from cfg_select itself, it comes from the cfg_select's input, unmodified.
For macros like this it is the correct behavior to keep the input token spans (together their editions, macro backtraces, etc.), imagine if e.g. derive, which behaves similarly, marked the item it's applied to as macro expanded when re-emitting it.
For cfg_select the behavior is also consistent with the regular cfg.

[blyxyas]

In that case, we should mark it in some other way :/. It's currently a lint, and I can imagine other lints breaking because they're not equipped to handle it.

In the example of derive, we could have something similar to is_automatically_derived (be it either with a new DesugaringKind, or with a new attribute).

[bjorn3]

#[cfg_trace] like for #[cfg] or are there cases where cfg_select! can't apply atrributes to the output?

[estebank]

#[cfg_trace] like for #[cfg] or are there cases where cfg_select! can't apply atrributes to the output?

We didn't do that until recently: #157218

[saberoueslati]

@rustbot claim