rustdoc: use anonymous constant for primitives/keywords/attribute docs

[mejrs]

For context: #t-rustdoc > hiding fake_item modules. @ 💬

r? @GuillaumeGomez

I've rendered the standard library docs and prims/kws/attrs all show up as normal.

However I could not figure out how to run some of the tests, so this may be a bit of CI-debugging for me :(

[rustbot]

Some changes occurred in compiler/rustc_attr_parsing

cc @jdonszelmann, @JonathanBrouwer

Some changes occurred in compiler/rustc_passes/src/check_attr.rs

cc @jdonszelmann, @JonathanBrouwer

Some changes occurred in GUI tests.

cc @GuillaumeGomez

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 @obi1kenobi

[camelid]

However I could not figure out how to render std with private/hidden items shown (rustdocflags are ignored), nor how to run some of the tests, so this may be a bit of CI-debugging for me :(

Should be RUSTDOCFLAGS="--document-private-items --document-hidden-items" x doc library --stage 1 -- did that not work for you?

[mejrs]

However I could not figure out how to render std with private/hidden items shown (rustdocflags are ignored), nor how to run some of the tests, so this may be a bit of CI-debugging for me :(

Should be RUSTDOCFLAGS="--document-private-items --document-hidden-items" x doc library --stage 1 -- did that not work for you?

Hmm, I did not pass stage 1 before, but that indeed works,.

[camelid]

Without --stage 1, it might be using the stage0 rustdoc to document std, which will be without your changes.

[mejrs]

Without --stage 1, it might be using the stage0 rustdoc to document std, which will be without your changes.

Right, but I changed std to use anon consts so that should have errored . Maybe I don't quite understand how rustdoc interacts with bootstrapping. Or In just missed it before, we don't have that many private root modules in std actually. Regardless it works :)

[GuillaumeGomez]

Let's use the power of matches! =D

Suggested change

	if let Some(ItemKind::Const(ident, _gen, _ty, _rhs)) = item_kind
	&& ident.name == kw::Underscore
	{
	} else {
	if !matches!(item_kind, ItemKind::(ident, ..) if ident.name == kw::Underscore) {

View changes since the review

[camelid]

I think the old code is much clearer actually. Having a guard in the matches and then negating it is way more confusing IMO.

[GuillaumeGomez]

Please add a regression test documenting private items when these attributes (on consts) are used to ensure the const are not visible (you did it for json already, so that's a good start :3).

[rustbot]

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

[mejrs]

(Force pushed to remove some useless checks for "this fake module doesn't show up in html")

Please add a regression test documenting private items when these attributes (on consts) are used to ensure the const are not visible

Aren't const _ never visible? 🤔

In attribute parsing, we check that doc(attr/kw) can only be on constants named _. There is no such check for primitive docs but I would like in another PR to get rid of #[rustc_doc_primitive] and switch to #[doc(primitive = ...)], and there implement that check for primitives as well, if that's OK.

(you did it for json already, so that's a good start :3).

But that's to test that these prim/kw/attr fake items themselves never end up in rustdoc-json, it was never only about the modules.