rustdoc-json: Add test for #[optimize(..)]

[aDotInTheVoid]

Follow up to #138291

CC @jdonszelmann

r? @GuillaumeGomez

[rustbot]

These commits modify tests/rustdoc-json.
rustdoc-json is a public (but unstable) interface.

Please ensure that if you've changed the output:

• It's intentional.
• The FORMAT_VERSION in src/librustdoc-json-types is bumped if necessary.

cc @aDotInTheVoid, @obi1kenobi

[jdonszelmann]

What makes the test for this attr specifically useful? Since a bunch of them are changing which don't necessarily have tests either?

[jdonszelmann]

Don't get me wrong, not against more tests, was just wondering about specifically this one

[aDotInTheVoid]

I wrote it while working on rust-lang/rustdoc-types@33243c4, and wanted to see what exactly the difference was. It will also be useful for when #141358 happens.

[GuillaumeGomez]

Is the display of the inner value supposed to be like that? Shouldn't it be speed?

[aDotInTheVoid]

It's the debug printing of attributes, which mirrors their semantic strucure in HIR, not their semantic structure.

See https://doc.rust-lang.org/nightly/nightly-rustc/rustc_attr_data_structures/enum.OptimizeAttr.html and https://doc.rust-lang.org/nightly/nightly-rustc/rustc_attr_data_structures/enum.AttributeKind.html#variant.Optimize

[aDotInTheVoid]

See also

rust/tests/rustdoc-json/attrs/inline.rs

Lines 1 to 3 in e4b9d01

	//@ is "$.index[?(@.name=='just_inline')].attrs" '["#[attr = Inline(Hint)]"]'
	#[inline]
	pub fn just_inline() {}

[GuillaumeGomez]

I see. How attributes are displayed is incoherent though, which isn't great. For example:

attrs/repr_align.rs:3://@ is "$.index[?(@.name=='Aligned')].attrs" '["#[repr(align(4))]"]'
attrs/export_name_2024.rs:7://@ is "$.index[?(@.name=='example')].attrs" '["#[export_name = \"altered\"]"]'
attrs/repr_c.rs:13://@ is "$.index[?(@.name=='ReprCUnion')].attrs" '["#[repr(C)]"]'
attrs/repr_packed.rs:6://@ is "$.index[?(@.name=='Packed')].attrs" '["#[repr(packed(1))]"]'
keyword_private.rs:8://@ is   "$.index[?(@.name=='foo')].attrs" '["#[doc(keyword = \"match\")]"]'

Some of them are "code-like" and other represent the internal type representation in rustc, which is subject to change. Do we plan to uniform it at some point?

[aDotInTheVoid]

Yes. The code-like ones are either unstructured attributes, or have special handling in rustdoc.

#142936 moves to rustdoc-json not representing attributes as strings, but with a enum.

As part of #131229, all attributes are moving to being structured, but this is a gradual process.

[GuillaumeGomez]

Went through #142936. Feel very relieved. Attributes are declared in rustdoc-json-types so they can't change unexpectedly. Very good idea!

[GuillaumeGomez]

Especially in this case.

[GuillaumeGomez]

Thanks!

@bors r+ rollup

[bors]

📌 Commit bec2679 has been approved by GuillaumeGomez

It is now in the queue for this repository.