Hello Serde Team,

I've encountered a compile-time panic when using #[derive(Deserialize)] in combination with #[serde(rename_all = "camelCase")] on a struct that has fields with non-ASCII identifiers (specifically, Chinese characters).

Summary

The serde_derive procedural macro appears to panic during compilation when it tries to apply the camelCase renaming rule to a field name containing multi-byte UTF-8 characters. The error message clearly indicates a byte-indexing issue on a non-char boundary, suggesting the renaming logic might be incorrectly assuming ASCII strings.

The issue does not occur if:

• The #[serde(rename_all = "camelCase")] attribute is removed.
• The attribute is changed to #[serde(rename_all = "PascalCase")].

This suggests the bug is specific to the implementation of the camelCase transformation logic.

Minimal Reproducible Example (MRE)

Here is the smallest possible Rust project that demonstrates the bug:

Cargo.toml:

[package]
name = "serde-cjk-bug-report"
version = "0.1.0"
edition = "2024"

[dependencies]
serde = { version = "1.0.219", features = ["derive"] }

src/main.rs:

use serde::Deserialize;

// This struct definition causes the compiler to panic.
#[derive(Deserialize, Debug, Clone)]
#[serde(rename_all = "camelCase")]
pub struct TestPayload {
    // Field name with Chinese characters
    pub 项目名称: String,
    pub 项目地址: String,
}

fn main() {
    // The panic happens at compile time, so the main function's content
    // doesn't matter, but it's here to make the example complete.
    println!("This code will not compile.");
}

Compiler Panic Output

When trying to compile the MRE with cargo build, the compiler panics with the following output:

error: proc-macro derive panicked
  --> src/main.rs:5:10
   |
5  | #[derive(Deserialize, Debug, Clone)]
   |          ^^^^^^^^^^^
   |
   = help: message: byte index 1 is not a char boundary; it is inside '项' (bytes 0..3) of `项目名称`

Expected vs. Actual Behavior

• Expected Behavior: The code should compile successfully. The camelCase rule, when applied to a field name like 项目名称 which has no case or separators to modify, should ideally result in no change to the name, or at least not cause a panic.
• Actual Behavior: The compiler panics with a fatal error related to string indexing.

Analysis

The panic message byte index 1 is not a char boundary strongly suggests that the string manipulation logic within the rename_all = "camelCase" implementation is incorrectly treating the UTF-8 field identifier as a sequence of bytes rather than characters. It attempts a byte-level index access, which is unsafe for multi-byte strings. A correct implementation should iterate over chars() rather than bytes().

Workarounds

For others encountering this issue, the current workarounds are:

1. Avoid using non-ASCII field names if rename_all is needed.
2. Use ASCII field names and the #[serde(rename = "项目名称")] attribute on each field individually.
3. Remove the rename_all attribute if the exact casing is not critical for your use case.

Environment

• Rust Version: rustc 1.89.0 (29483883e 2025-08-04)
• Serde Version: 1.0.219
• OS: Windows 10

Thank you for maintaining this essential crate! I hope this report is helpful in resolving the issue.