Regression from new lint out_of_scope_macro_calls

[tgross35]

compiler_builtins is now getting:

 error: cannot find macro `define_rust_probestack` in this scope
   --> src/probestack.rs:134:5
    |
137 |     define_rust_probestack!(
    |     ^^^^^^^^^^^^^^^^^^^^^^
    |
    = warning: this was previously accepted by the compiler but is being phased out; it will become a hard error in a future release!
    = note: for more information, see issue #124535 <https://github.com/rust-lang/rust/issues/124535>
    = help: import `macro_rules` with `use` to make it callable above its definition
    = note: `-D out-of-scope-macro-calls` implied by `-D warnings`
    = help: to override `-D warnings` add `#[allow(out_of_scope_macro_calls)]`

But this doesn't seem right, unless I am misreading something. The global_asm! invocation that is enabled on x86_64 (https://github.com/rust-lang/compiler-builtins/blob/7a84ce4a90cd2d22494daead50f9026b5ca52140/src/probestack.rs#L130-L134) should find the define_rustc_probestack definition that is enabled on everything except apple or uefi (https://github.com/rust-lang/compiler-builtins/blob/7a84ce4a90cd2d22494daead50f9026b5ca52140/src/probestack.rs#L61-L79).

CI: https://github.com/rust-lang/compiler-builtins/actions/runs/9675571819/job/26693340492?pr=636#step:5:24

Unfortunately I haven't been successful in finding a mcve, https://play.rust-lang.org/?version=nightly&mode=debug&edition=2021&gist=f299db7ba55daaa5b7721323bf72cfb8 works fine.

[tgross35]

I can reproduce locally with rustc 1.81.0-nightly (fda509e 2024-06-25)

@rustbot label +E-needs-mcve +A-lint +T-compiler +regression-from-stable-to-nightly -needs-triage

Cc @petrochenkov since you added the lint in #125741

[petrochenkov]

Yes, looks like a false positive caused by global_asm! needing a special treatment.

[petrochenkov]

It's not about global asm, it's about any eager expansion in macro call items.
Fixed in #126987.

[tgross35]

Awesome, thanks for the quick fix

[tgross35]

There might be another bug with this lint, seems like it might not be recognized as an inner attribute. Trying to ignore it gives contradictory messages:

error: unknown lint: `out_of_scope_macro_calls`
  --> src/probestack.rs:52:10
   |
52 | #![allow(out_of_scope_macro_calls)]
   |          ^^^^^^^^^^^^^^^^^^^^^^^^
   |
   = note: `-D unknown-lints` implied by `-D warnings`
   = help: to override `-D warnings` add `#[allow(unknown_lints)]`

error: cannot find macro `define_rust_probestack` in this scope
   --> src/probestack.rs:265:5
    |
265 |     define_rust_probestack!(
    |     ^^^^^^^^^^^^^^^^^^^^^^
    |
    = warning: this was previously accepted by the compiler but is being phased out; it will become a hard error in a future release!
    = note: for more information, see issue #124535 <https://github.com/rust-lang/rust/issues/124535>
    = help: import `macro_rules` with `use` to make it callable above its definition
    = note: `-D out-of-scope-macro-calls` implied by `-D warnings`
    = help: to override `-D warnings` add `#[allow(out_of_scope_macro_calls)]`

From trying to silence the error in compiler_builtins, https://github.com/rust-lang/compiler-builtins/actions/runs/9735325154/job/26864417423#step:13:2321

[beetrees]

I don't think the lint is ever actually being registered anywhere; more specifically it needs to be added to the HardwiredLints lint pass at the top of rustc_lint_defs/src/builtin.rs.