False positive `dead_code` lint when type is only used via destructuring

[daniel5151]

Code

struct Foo(usize);

fn get_thing<T>() -> T { todo!() }

pub fn foo() {
    let Foo(_x) = get_thing();
    
    // no warning if the type is explicitly used when calling get_thing
    // let Foo(_x) = get_thing::<Foo>();
}

Current output

warning: struct `Foo` is never constructed
 --> src/lib.rs:1:8
  |
1 | struct Foo(usize);
  |        ^^^
  |
  = note: `#[warn(dead_code)]` on by default

Desired output

(nothing)

Rationale and extra context

For context, the specific "get_thing" function I originally ran into this was with zerocopy::FromBytes::read_from_bytes.

Other cases

Rust Version

$ rustc --version --verbose
rustc 1.82.0 (f6e511eec 2024-10-15)
binary: rustc
commit-hash: f6e511eec7342f59a25f7c0534f1dbea00d01b14
commit-date: 2024-10-15
host: x86_64-unknown-linux-gnu
release: 1.82.0
LLVM version: 19.1.1

Anything else?

No response

[m-ysk]

@rustbot claim

[m-ysk]

@daniel5151
I've submitted a PR #133128 to resolve this issue and the reviewer thinks that the code you made as an MRE should be warned as a dead code as you can see in the following link.

#133128 (comment)

I also agree to the reviewer, but at the same time I'm thinking that there's a possibility that you've actually got a bug, but you've over-generalized your code when creating the above MRE because you said that you've found this issue when using zerocopy::FromBytes::read_from_bytes.
So, if possible, could you give me another MRE closer to your original code by using the zerocopy crate?

[daniel5151]

Sure. Here's a repro using zerocopy which is closer to the real (and IMO, totally reasonable) code I was writing:

use zerocopy; // 0.7.35
use zerocopy_derive; // 0.7.35

use zerocopy::FromBytes;

#[derive(zerocopy_derive::FromZeroes, zerocopy_derive::FromBytes)]
#[repr(C, packed)]
struct PacketHeader {
    pub kind: u8,
    pub len: u32
}

fn main() {
    let packet = [0x1, 2, 0, 0, 0, b'h', b'i'];
    let PacketHeader { kind, len } = FromBytes::read_from_prefix(&packet).unwrap();
    dbg!((kind, len));
}

Output:

   Compiling playground v0.0.1 (/playground)
warning: struct `PacketHeader` is never constructed
 --> src/main.rs:8:8
  |
8 | struct PacketHeader {
  |        ^^^^^^^^^^^^
  |
  = note: `#[warn(dead_code)]` on by default

warning: `playground` (bin "playground") generated 1 warning
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.48s
     Running `target/debug/playground`
[src/main.rs:16:5] (kind, len) = (
    1,
    2,
)

[m-ysk]

Thanks. Your new repro seems reasonable to me. I'll think about how to go ahead with my PR #133128 again.