proc macro attributes considered to be `unresolvedReference`

[jhgg]

consider the following code used w/ the serde library:

use serde::Deserialize;

#[derive(Deserialize)]
struct Foo {
    #[serde(default)]
   // ^^^^^ unresolvedReference 
    bar: u64,
}

We should be able to figure out if this is a valid attribute or not, by looking at the things the structs derive from and looking at the attributes for each proc macro to compute a list of identifiers that are valid within this context:

// in serde_derive/src/lib.rs
#[proc_macro_derive(Deserialize, attributes(serde))]
                                         // ^^^^^ we should find these.
pub fn derive_deserialize(input: TokenStream) -> TokenStream {
    let mut input = parse_macro_input!(input as DeriveInput);
    de::expand_derive_deserialize(&mut input)
        .unwrap_or_else(to_compile_errors)
        .into()
}

Bonus points would be to have a hover implementation that shows which proc macro's are consuming this attribute as well!

rust-analyzer version: 6434ada 2021-12-05 dev
rustc-version: 1.57.0

[Veykril]

Ah right, I completely forgot about (derive) helper attributes which we now show as unresolved since they aren't being mapped to anything by macros...

As you said we can do some specific things here to fix that for derive helpers specifically, but the problem persists for helper attributes of proc-macro attributes as well like with salsa here

In which case we can't know about these

So I guess we'll have to add that attribute highlight heuristics back for now...

I do like the hover idea though 👍

[Veykril]

Actually now I wonder if it would make sense to push proc macro authors to expose dummy attributes that resolve to these helpers. That way people could also hover them for more info on the helper itself 🤔

[Veykril]

Opened a zulip thread for the latter idea.

Highlighting issue papered over again with #10937 for now.