Description
I'm currently writing a crate where I'd like to use doc comments as attributes for a proc macro that generates context directives for errors and log messages via tracing
.
https://github.com/yaahc/spandoc/blob/master/src/lib.rs
I had originally assumed that if my proc macro stripped out the doc attribute after converting it to a span then the compiler would not lint on the doc comments that actually got used as code, but when I tested this it still lints which implies to me that the unused doc comment lint happens in an earlier pass than proc-macro expansion does.
I'd like to be able to avoid using #[allow(unused_doc_comments)]
on either the whole function or file because I want the lints to help me find doc comments that I add, expecting them to turn into spans, but which my proc macro does not handle and therefore leaves in as doc comments.