Special case for empty comments in doc comment kind

[edwin0cheng]

Part of #4103

Fix ui/empty/empty-comment.rs macros

[matklad]

bors r+

I wonder though, if we can/need fix something on the macro side as well.. It seems weird that macro non-panicing depends on an impl details of a function elsewhere. Perhaps we can lift some invariants to types or something...

[bors]

Build failed:

• Rust (macos-latest)

[edwin0cheng]

I wonder though, if we can/need fix something on the macro side as well.. It seems weird that macro non-panicing depends on an impl details of a function elsewhere. Perhaps we can lift some invariants to types or something...

The panic is actually raised by here:

https://github.com/rust-analyzer/rust-analyzer/blob/f654f4943552888c2b8f079cf687970268d81f18/crates/ra_mbe/src/syntax_bridge.rs#L164-L177

It bascially try to convert a comment text to #[doc(sth)] , I think we could move this logic to ast::make ?

[edwin0cheng]

bors retry

[bors]

Build succeeded:

• Rust (macos-latest)
• Rust (ubuntu-latest)
• Rust (windows-latest)
• TypeScript

[matklad]

So, I think semantically what we do here is that we extract contents of the comment. We have DocCommentsOwner::doc_comment_text which does a similar job. I probably would expect an ast::Comment::contents() -> String method to be used to implement both getting docs (which is stiched form many line comments) and for lowering doc comemnts to attributes.