--quality-delta reports Drop::drop as dead code: the compiler calls it at scope end #494

ripwire: 0.6.9 (Release, GNU 14.2.1)
Verb: --quality-delta

What happens

impl Drop for T { fn drop(&mut self) { … } } is called by the compiler when a T goes out of scope. No source line names drop, so the index has no edge to it, and the method is listed as dead code.

Minimal reproduction

W=$(mktemp -d) && mkdir -p "$W/src" && cd "$W" && git init -q .
printf '[package]\nname = "repro"\nversion = "0.1.0"\nedition = "2021"\n\n[dependencies]\n' > Cargo.toml
printf 'pub fn api() -> u32 {\n    2\n}\n' > src/lib.rs
git add -A && git -c user.name=r -c user.email=r@example.invalid commit -qm base
cat > src/lib.rs <<'RS'
pub fn api() -> u32 {
    use_guard()
}

struct Guard(u32);

impl Drop for Guard {
    fn drop(&mut self) {
        self.0 = 0;
    }
}

fn use_guard() -> u32 {
    let guard = Guard(3);
    guard.0
}
RS
cargo build
ripwire . --quality-delta --legend=compact

ripwire prints one row:

<r kind="dead-code" sym="src/lib.rs::Guard::drop" origin="new-symbol" p="src/lib.rs:8"/>

use_guard is not reported, because api calls it.

Control

A plain method that nobody calls would be reported, and it should be. The difference is that the caller of Drop::drop is implicit, and nothing in the index records it.

Why it matters

A guard type that cleans up in Drop is the usual way to make cleanup run on every exit path, including when a future is cancelled. The obvious workaround, an explicit cleanup call, does not run when the future is dropped mid-way. So the row cannot be fixed on the user side without a weaker design. It does not gate, but it shows up on every project that uses a cleanup guard.

Expected

drop in an impl Drop for … block counts as called, the way Python hooks and framework entry points are already excluded from --quality-delta's dead-code rows (PR #437, closing #228).