1.38 regression ? proc-macro derive panicked if there is r# in front of the attribute name

[fdubois1]

I have a macro and it defines an attribute called "async". When I use it, I add r# in front of it. I have this :

#[cowrpc(r#async)]
struct Handshake {
...
}

Since 1.38.0, it fails with

error: proc-macro derive panicked
 --> src/cow/den_server_iface.rs:3:10
  |
3 | #[derive(CowRpcIface)]
  |          ^^^^^^^^^^^
  |
  = help: message: "r#async" is not a valid ident

error: proc-macro derive panicked

It was building without any error before. What is different with 1.38?

[petrochenkov]

Looks very similar to #63894 (pretty-printing change making raw identifiers to be printed precisely).

[petrochenkov]

I suspect the attribute or derive expects identifier async in its input, but doesn't expect r#async.

(It would be nice if someone could check how the expanded code looks like and whether my guess is correct.)

[fdubois1]

@petrochenkov How can I see what the expanded code looks like ?

Thanks

[petrochenkov]

cargo expand is able to show expanded code for whole crates with dependencies.

[fdubois1]

FYI, here is the macro: https://github.com/Devolutions/cowrpc-rs/tree/master/cowrpc_derive/src

I looked the code generated in my case. I'm not sure to understand your point when you say that it expects identifier async in its input. I looked the macro code (https://github.com/Devolutions/cowrpc-rs/tree/master/cowrpc_derive/src), it is true that it compare the ident name with "async" (https://github.com/Devolutions/cowrpc-rs/blob/master/cowrpc_derive/src/iface/attr.rs#L30), but it would raise an error if you provide something that macro doesn't support, for example #[cowrpc(foostring)] would raise an error "There is no word attribute you can use on data types with cowrpc".

I can see that we use syn: 0.13 and proc-macro2: 0.3. I guess the error message is comming at the beginning when we do syn::parse(input)... but if we didn't change the dependencies, should not be a problem...

FYI, I updated the code to edition 2018 where async is now a reserved word. It changed code to add lot of r# in front of async, but should not be a problem.

Thanks for your help

[o0Ignition0o]

@rustbot modify labels to +I-prioritize

[LeSeulArtichaut]

Assigning P-medium as discussed as part of the Prioritization WG procedure.