This is awesome, but I don't know enough about HIR to review I think. Is there a way to test this somehow? Maybe you could see if #80414 now generates documentation at core/marker/macro.Copy.html instead of core/macro.Copy.html ? See https://rustc-dev-guide.rust-lang.org/rustdoc-internals.html#dotting-is-and-crossing-ts for how to write rustdoc tests.

I can't get such a rustdoc test passing.

Ok, then it looks like there were two separate bugs for #74355, one in rustc and one in rustdoc.

r? @petrochenkov maybe?

There is still some effort needed to fix #74355, but the changes in this PR are great since they will fix the hack I have used with my fix attempt in #77862

What tcx.hir().parent_module(hir_id) and tcx.parent(def_id) show for #[macro_export] macro_rules! ... items now?

If they are consistent (looks like they should be), then this this PR is good.
However, rustdoc may start showing #[macro_export] macro_rules! ... in inner modules rather than in the crate root, in that case it will need to be fixed.

@cjgillot Could you check both these cases? If they behave as expected, then r=me.

@petrochenkov, I am not sure how to address your review. How do you recommend I test the consistency?

@cjgillot
In any way, e.g. printf from the compiler.
I don't suggest to put this check into test suite, just look at the IDs and make sure that they are consistent.

@petrochenkov, ok thanks.

I tested adding the following in the check_attr HIR visitor, and tested a file with both top-level and module macro exports. It passes.

use rustc_middle::ty::DefIdTree;
let def_id = self.tcx.hir().local_def_id(macro_def.hir_id);
assert_eq!(
    Some(self.tcx.parent_module(macro_def.hir_id).to_def_id()),
    self.tcx.parent(def_id.to_def_id())
);

Should be good to go.

@bors r=petrochenkov

📌 Commit 4fb1236 has been approved by petrochenkov

⌛ Testing commit 4fb1236 with merge 41601ef...

☀️ Test successful - checks-actions
Approved by: petrochenkov
Pushing 41601ef to master...