Hygiene for macro-generated newtype struct deserialization with with attr
#2847
Conversation
error[E0425]: cannot find value `__e` in this scope
--> test_suite/tests/regression/issue2846.rs:12:19
|
12 | declare_in_macro!("with");
| ^^^^^^ not found in this scope
warning: field `0` is never read
--> test_suite/tests/regression/issue2846.rs:8:45
|
8 | pub struct S(#[serde(with = $with)] i32);
| - field in this struct ^^^
...
12 | declare_in_macro!("with");
| ------------------------- in this macro invocation
|
= help: consider removing this field
= note: `#[warn(dead_code)]` on by default
= note: this warning originates in the macro `declare_in_macro` (in Nightly builds, run with -Z macro-backtrace for more info)
|
This seems very footgunny, do you know if there is an issue or discussion about making this less painful? It feels to me like a wrong-defaults issue. I'd expect it to always use That said, wouldn't it be cleaner to do that manually here and thus also avoid clippy issues? |
|
- quote_spanned!(serialize_with.span()=> {
+ quote_spanned!(serialize_with.span().resolved_at(Span::call_site())=> {
#[doc(hidden)]
struct __SerializeWith #wrapper_impl_generics #where_clause {has these unwanted effects, among others: error[E0277]: the trait bound `&u8: Serializer` is not satisfied
--> tests/ui/with/incorrect_type.rs:14:10
|
-14 | #[derive(Serialize, Deserialize)]
- | ^^^^^^^^^ the trait `Serializer` is not implemented for `&u8`
-15 | struct W(#[serde(with = "w")] u8, u8);
- | --- required by a bound introduced by this call
+14 | #[derive(Serialize, Deserialize)]
+ | ^--------
+ | |
+ | __________in this derive macro expansion
+ | |
+15 | | struct W(#[serde(with = "w")] u8, u8);
+ | | --^
+ | |_________________________|_|
+ | | the trait `Serializer` is not implemented for `&u8`
+ | required by a bound introduced by this call
|
= help: the following other types implement trait `Serializer`:
&mut Formatter<'a> error[E0061]: this function takes 1 argument but 2 arguments were supplied
--> tests/ui/with/incorrect_type.rs:15:25
|
15 | struct W(#[serde(with = "w")] u8, u8);
| ^^^ unexpected argument #2 of type `__S`
|
note: function defined here
--> tests/ui/with/incorrect_type.rs:9:12
|
9 | pub fn serialize<T, S: Serializer>(_: S) -> Result<S::Ok, S::Error> {
| ^^^^^^^^^ ----
+help: remove the extra argument
+ |
+15 - struct W(#[serde(with = "w")] u8, u8);
+15 + struct W(#[serde(with = )] u8, u8);
+ | |
|
Hmm... if we can make the world of proc macros better by stopping to do that, that seems great. I was working on some span stuff just a few months ago. I'll see if we can remove the PartialEq impl from spans entirely, it's already a footgun just in the compiler, seems best to remove it if its "benign" uses are problematic, for proc macros |
Fixes #2846.