`non_local_definitions` triggers on macro-generated code

[Nemo157]

I tried this code:

> cargo new foo && cd foo
    Creating binary (application) `foo` package
note: see more `Cargo.toml` keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html

> cargo add displaydoc
    Updating crates.io index
      Adding displaydoc v0.2.4 to dependencies
             Features:
             + std
    Updating crates.io index
     Locking 6 packages to latest compatible versions

> echo '#[derive(displaydoc::Display)] pub enum Foo { #[doc = "test"] Test }' >> src/main.rs

I expected to see this happen: no warnings

Instead, this happened:

> cargo check
    Checking foo v0.1.0 (/tmp/scratch.rust.2024-05-28T21-14.UGSjii/foo)
warning: non-local `impl` definition, they should be avoided as they go against expectation
 --> src/main.rs:4:10
  |
4 | #[derive(displaydoc::Display)] pub enum Foo { #[doc = "test"] Test }
  |          ^^^^^^^^^^^^^^^^^^^
  |
  = help: move this `impl` block outside the of the current constant `_DERIVE_Display_FOR_Foo`
  = note: an `impl` definition is non-local if it is nested inside an item and may impact type checking outside of that item. This can be the case if neither the trait or the self type are at the same nesting level as the `impl`
  = note: one exception to the rule are anon-const (`const _: () = { ... }`) at top-level module and anon-const at the same nesting as the trait or type
  = note: this lint may become deny-by-default in the edition 2024 and higher, see the tracking issue <https://github.com/rust-lang/rust/issues/120363>
  = note: the derive macro `displaydoc::Display` may come from an old version of the `displaydoc` crate, try updating your dependency with `cargo update -p displaydoc`
  = note: `#[warn(non_local_definitions)]` on by default
  = note: this warning originates in the derive macro `displaydoc::Display` (in Nightly builds, run with -Z macro-backtrace for more info)

warning: `foo` (bin "foo") generated 1 warning
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.09s

Meta

rustc --version --verbose:

rustc 1.80.0 (84b40fc90 2024-05-27)
binary: rustc
commit-hash: 84b40fc908c3adc7e0e470b3fbaa264df0e122b8
commit-date: 2024-05-27
host: x86_64-unknown-linux-gnu
release: 1.80.0
LLVM version: 18.1.6

cc #120363

[Nemo157]

It looks like there is an upstream issue about this yaahc/displaydoc#46, but it seems like this shouldn't be an issue for dependents as long as displaydoc is on edition < 2024.

[Urgau]

It is expected for the non_local_definitions lint to lint on macro-generated code.
It is also expected for the lint to lint on all editions (it is not a edition dependent lint).

[Nemo157]

What am I meant to do as a downstream of a macro crate that hasn't fixed it then? I don't want to be #[allow]ing a FCW, but there's nothing else I can do to clear this warning from my code. (And that implies that the planned "hard error in 2027" actually means for all editions, not just Edition 2027?)

[Urgau]

What am I meant to do as a downstream of a macro crate that hasn't fixed it then?

Allow the lint. Report the issue upstream. Consider switching to an unaffected crate.

I don't want to be #[allow]ing a FCW

That's understandable, but until the upstream crate is fixed I don't have a better solution.

Note, that it is quite unlikely that this point that the lint will become deny-by-default in Rust 2024.

(And that implies that the planned "hard error in 2027" actually means for all editions, not just Edition 2027?)

No, if hard-error there is, it would only be for edition 2027 and higher, not below.

[workingjubilee]

It is expected for the non_local_definitions lint to lint on macro-generated code.

Why?

[workingjubilee]

Detecting and ignoring macro-generated code for a lint where possible is the expected behavior of rustc.

[Urgau]

Because, in particular for macro-generated code, we can't lint at the source since, well it is only generated when using the proc-macro or macro_rules!.

Detecting and ignoring macro-generated code for a lint where possible is the expected behavior of rustc.

Yes, but we must still emit the warning in macro-generated code, since the goal is to make the lint deny-by-default and then hard-error in a future edition, and without the warn it would go straight to hard-error which is not good.

[workingjubilee]

Because, in particular for macro-generated code, we can't lint at the source since, well it is only generated when using the proc-macro or macro_rules!.

macro authors write tests.

[workingjubilee]

Is there no ability to enable the lint when in the same crate/workspace but disable it for "foreign" macro-generated spans?