rustdoc: Use DiagnosticInfo in more parts of intra-doc links #83875
Conversation
jyn514
added
C-cleanup
|
@bors delegate=bugadani |
|
✌️ @bugadani can now approve this pull request |
This comment has been minimized.
This comment has been minimized.
There was a problem hiding this comment.
Most of this is straight forward unless I missed a subtle change somewhere where the transformation is not equivalent.
IIRC when we were working on this part, I was considering going all-in with DiagnosticInfo but I think I didn't feel comfortable enough with the change in L1121 (the one I commented on). I think that one might need some double checking.
I think there might still exist a slight inconsistency as DiagnosticInfo is passed by value in most of the places, but by ref in a few others. No big deal though IMO.
| path_str, | ||
| dox, | ||
| ori_link.range, | ||
| diag_info, |
There was a problem hiding this comment.
I think this might be a slight change in behaviour as path_str != ori_link.link. I don't remember the details but if this doesn't trip a test, maybe we should have one that covers it?
There was a problem hiding this comment.
It's a bug to ever pass in path_str here, otherwise it won't show the anchor at the end. Here's an example with the old diagnostic:
warning: `usize` contains an anchor, but links to builtin types are already anchored
--> anchor.rs:1:6
|
1 | //! [prim@usize#x]
| ^^^^^^^^^^^^ contains invalid anchor
|
= note: `#[warn(rustdoc::broken_intra_doc_links)]` on by default
I'll add a test.
FWIW, the reason I pass path_str separately from diag_info in some places is exactly so the caller isn't responsible for remembering to update it.
Yeah, I ran into some borrow checker issues ... I really wish there were some fix for rust-lang/rfcs#2848. |
ori_link contains anchors, path_str does not. It's important that anchor_failure be passed a link with the anchors still present.
This gets rid of a lot of parameters, as well as fixing a diagnostic bug.
It's not the range of the full link, it's only a partial range.
| @@ -43,3 +43,7 @@ pub fn enum_link() {} | |||
| /// [u32#hello] | |||
| //~^ ERROR `u32#hello` contains an anchor | |||
| pub fn x() {} | |||
|
|
|||
| /// [prim@usize#x] | |||
There was a problem hiding this comment.
Actually, why is this an error? Shouldn't this link to an anchor in a page like primitive.usize.html? I guess this also goes for the test case above ([u32#hello]). I think I would find a link to usize#examples-3 actually reasonable. What am I missing?
There was a problem hiding this comment.
This is #83083, it's a bug but pretty hard to fix.
|
@bors r+ |
|
📌 Commit a86a740 has been approved by |
bors
added
S-waiting-on-bors
rustdoc: Use DiagnosticInfo in more parts of intra-doc links This makes the code a lot less verbose. This is separated into lots of tiny commits because it was easier for me that way, but the overall diff isn't that big if you want to read it at once. r? `@bugadani`
|
☀️ Test successful - checks-actions |
This makes the code a lot less verbose.
This is separated into lots of tiny commits because it was easier for me that way, but the overall diff isn't that big if you want to read it at once.
r? @bugadani