[WIP] Distinguish between links on inner and outer attributes

[jyn514]

Normally, rustdoc will resolve //! docs in the scope inside the module
and /// in the scope of the parent. But if there are modules with
inner docs, it would previously resolve all of the links inside the module.

This now distinguishes between links that came from inside the module
and links that came from outside. The intra-doc links part works, but
unfortunately the rest of rustdoc assumes there is only one canonical
resolution for a link and won't look for more than one on the same item.

• Store the attribute style on each DocFragment
• Don't combine attributes with different styles
• Calculate the parent module in only one place
• Remove outdated and fixed FIXME comments

This came from a failed attempt to work on #78591 and there's some debugging for that left over. Let me know if I should remove it.

r? @GuillaumeGomez - do you have suggestions for how to pass this information to the rest of rustdoc? For context, the issue is that given a list of links like

vec![ItemLink { link: "a", did: Some(DefId(0)) }, ItemLink { link: "a", did: Some(DefId(1)) }]

rustdoc::html will always pick the first resolution in the array. Maybe the link should also say whether it's an inner or outer link, and have html distinguish too?

cc @Manishearth - technically this is a breaking change, but I consider the old behavior a bug.

[GuillaumeGomez]

Question: wouldn't it be simpler to first treat inner docs and then outer docs for each item? Therefore, you could change the parent_node depending on the type and makes your life a bit easier no? (and that could even allow to have a smaller fold_item function :p)

[jyn514]

first treat inner docs and then outer docs for each item

This is what it does currently, no? Or do you mean to have separate clean::Items for inner and outer attributes? That seems fragile, I don't know how the rest of rustdoc will react to duplicate items.

[GuillaumeGomez]

From what I understand, not exactly. We have a weird mix between the two and then treat them as one. What I meant was to literally try to split both handling so that the parent id isn't "special".

[jyn514]

This needs to also update parent_node or any following attributes will still use the outer scope.

Suggested change

	let base_node = if !seen_inner_docs && item.is_mod() && attr.style == AttrStyle::Inner {
	// FIXME(jynelson): once `Self` handling is cleaned up I think we can get rid
	// of `mod_ids` altogether
	self.mod_ids.push(item.def_id);
	seen_inner_docs = true;
	Some(item.def_id)
	} else {
	parent_node
	};
	if !seen_inner_docs && item.is_mod() && attr.style == AttrStyle::Inner {
	// FIXME(jynelson): once `Self` handling is cleaned up I think we can get rid
	// of `mod_ids` altogether
	self.mod_ids.push(item.def_id);
	seen_inner_docs = true;
	parent_node = Some(item.def_id);
	}

and then change base_node to parent_node below.

[bors]

☔ The latest upstream changes (presumably #78956) made this pull request unmergeable. Please resolve the merge conflicts.

Note that reviewers usually do not review pull requests until merge conflicts are resolved! Once you resolve the conflicts, you should change the labels applied by bors to indicate that your PR is ready for review. Post this as a comment to change the labels:

@rustbot modify labels: +S-waiting-on-review -S-waiting-on-author

[crlf0710]

Triage: CI is still red here.

[jyn514]

I'm not convinced this is actually useful ... It doesn't fix the ICE and a lot of the rest of rustdoc assumes this is the same. It might be easier to just forbid intra-doc-links to appear on both inner and outer attributes.

[jyn514]

I don't plan to follow up on this.