Incorrect trait bounds suggestion with derive macros

[Kixiron]

Given the following code:

Commit

The current output is:

 λ  cargo check
    Checking dbsp v0.1.0 (G:\Users\Chase\Code\Rust\database-stream-processor)
error[E0277]: the trait bound `R: Neg` is not satisfied
   --> src\trace\ord\indexed_zset_batch.rs:314:13
    |
314 |     result: <Layers<K, V, R, O> as Trie>::MergeBuilder,
    |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ the trait `Neg` is not implemented for `R`
    |
note: required for `R` to implement `IntoErasedDiff`
   --> src\trace\layers\erased\vtable.rs:227:9
    |
227 | impl<T> IntoErasedDiff for T
    |         ^^^^^^^^^^^^^^     ^
note: required for `TypedLayer<V, R>` to implement `Trie`
   --> src\trace\layers\erased\mod.rs:306:12
    |
306 | impl<K, R> Trie for TypedLayer<K, R>
    |            ^^^^     ^^^^^^^^^^^^^^^^
help: consider further restricting this bound
    |
310 |     R: DBWeight + std::ops::Neg,
    |                 +++++++++++++++

error[E0277]: the trait bound `for<'a> &'a R: Neg` is not satisfied
   --> src\trace\ord\indexed_zset_batch.rs:314:13
    |
314 |     result: <Layers<K, V, R, O> as Trie>::MergeBuilder,
    |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ the trait `for<'a> Neg` is not implemented for `&'a R`
    |
note: required for `R` to implement `NegByRef`
   --> src\algebra\mod.rs:168:9
    |
168 | impl<T> NegByRef for T
    |         ^^^^^^^^     ^
note: required for `R` to implement `IntoErasedDiff`
   --> src\trace\layers\erased\vtable.rs:227:9
    |
227 | impl<T> IntoErasedDiff for T
    |         ^^^^^^^^^^^^^^     ^
note: required for `TypedLayer<V, R>` to implement `Trie`
   --> src\trace\layers\erased\mod.rs:306:12
    |
306 | impl<K, R> Trie for TypedLayer<K, R>
    |            ^^^^     ^^^^^^^^^^^^^^^^
help: consider extending the `where` clause, but there might be an alternative better way to express this requirement
    |
311 |     O: OrdOffset, &'a R: for<'a> Neg
    |                 ~~~~~~~~~~~~~~~~~~~~

error[E0277]: the trait bound `R: Neg` is not satisfied
   --> src\trace\ord\indexed_zset_batch.rs:305:10
    |
305 | #[derive(SizeOf)]
    |          ^^^^^^ the trait `Neg` is not implemented for `R`
    |
note: required for `R` to implement `IntoErasedDiff`
   --> src\trace\layers\erased\vtable.rs:227:9
    |
227 | impl<T> IntoErasedDiff for T
    |         ^^^^^^^^^^^^^^     ^
note: required for `TypedLayer<V, R>` to implement `Trie`
   --> src\trace\layers\erased\mod.rs:306:12
    |
306 | impl<K, R> Trie for TypedLayer<K, R>
    |            ^^^^     ^^^^^^^^^^^^^^^^
    = note: this error originates in the derive macro `SizeOf` (in Nightly builds, run with -Z macro-backtrace for more info)
help: consider further restricting this bound
    |
310 |     R: DBWeight + std::ops::Neg,
    |                 +++++++++++++++

error[E0277]: the trait bound `for<'a> &'a R: Neg` is not satisfied
   --> src\trace\ord\indexed_zset_batch.rs:305:10
    |
305 | #[derive(SizeOf)]
    |          ^^^^^^ the trait `for<'a> Neg` is not implemented for `&'a R`
    |
note: required for `R` to implement `NegByRef`
   --> src\algebra\mod.rs:168:9
    |
168 | impl<T> NegByRef for T
    |         ^^^^^^^^     ^
note: required for `R` to implement `IntoErasedDiff`
   --> src\trace\layers\erased\vtable.rs:227:9
    |
227 | impl<T> IntoErasedDiff for T
    |         ^^^^^^^^^^^^^^     ^
note: required for `TypedLayer<V, R>` to implement `Trie`
   --> src\trace\layers\erased\mod.rs:306:12
    |
306 | impl<K, R> Trie for TypedLayer<K, R>
    |            ^^^^     ^^^^^^^^^^^^^^^^
    = note: this error originates in the derive macro `SizeOf` (in Nightly builds, run with -Z macro-backtrace for more info)
help: consider extending the `where` clause, but there might be an alternative better way to express this requirement
    |
305 | #[derive(SizeOf, &'a R: for<'a> Neg)]
    |                ++++++++++++++++++++

For more information about this error, try `rustc --explain E0277`.
error: could not compile `dbsp` due to 4 previous errors

Which has some weird suggestions, mainly #[derive(SizeOf, &'a R: for<'a> Neg)]. Additionally, the "alternative better way" of "but there might be an alternative better way to express this requirement" is worded a little strangely, it's kind of a double positive

 λ  rustc -Vv
rustc 1.66.0-nightly (7fcf850d7 2022-10-23)
binary: rustc
commit-hash: 7fcf850d7942804990a1d2e3fe036622a0fe4c74
commit-date: 2022-10-23
host: x86_64-pc-windows-msvc
release: 1.66.0-nightly
LLVM version: 15.0.2

 λ  cargo -Vv
cargo 1.66.0-nightly (071eeaf21 2022-10-22)
release: 1.66.0-nightly
commit-hash: 071eeaf210708219a5a1b2c4728ca2f97df7f2ae
commit-date: 2022-10-22
host: x86_64-pc-windows-msvc
libgit2: 1.5.0 (sys:0.15.0 vendored)
libcurl: 7.83.1-DEV (sys:0.4.55+curl-7.83.1 vendored ssl:Schannel)
os: Windows 10.0.19044 (Windows 10 Home) [64-bit]

[estebank]

The wording should likely have been "but there might be an alternative, better, way to express this requirement".

The suggestion machinery can check whether the span we're operating on is coming from a macro, but it might be difficult to differentiate between proc-macros that have properly annotated their spans so the suggestion works, and those that haven't (other than maybe try and compare span byte pos, depending on what we have available?).