Error when using unicode character literals

[paxbun]

bindgen panics when it encounters a character literal with a Unicode escape sequence representing a codepoint greater than 128.

Input C/C++ Header

#define OK 'a'
#define PANIC_CHAR_16 u'\uAC00'
#define PANIC_CHAR_32 U'\uAC00'

Bindgen Invocation

$ bindgen input.h -- -x c++
$ cargo run -p bindgen-cli -- input.h -- -x c++

Actual Results

thread 'main' panicked at 'assertion failed: `(left == right)`
  left: `3`,
 right: `1`', /Users/paxbun/.cargo/registry/src/github.com-1ecc6299db9ec823/bindgen-0.60.1/src/ir/var.rs:235:33
stack backtrace:
   0: _rust_begin_unwind
   1: core::panicking::panic_fmt
   2: core::panicking::assert_failed_inner
   3: core::panicking::assert_failed
   4: <bindgen::ir::var::Var as bindgen::parse::ClangSubItemParser>::parse
   5: <bindgen::ir::item::Item as bindgen::parse::ClangItemParser>::parse
   6: bindgen::parse_one
   7: bindgen::clang::visit_children
   8: __ZN5clang8cxcursor13CursorVisitor5VisitE8CXCursorb
   9: __ZN5clang8cxcursor13CursorVisitor25visitPreprocessedEntitiesINS_19PreprocessingRecord8iteratorEEEbT_S5_RS3_NS_6FileIDE
  10: __ZN5clang8cxcursor13CursorVisitor13VisitChildrenE8CXCursor
  11: _clang_visitChildren
  12: clang_sys::clang_visitChildren
  13: bindgen::Builder::generate
  14: std::panicking::try
  15: bindgen::main
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

Expected Results

/* automatically generated by rust-bindgen 0.64.0 */

pub const OK: u8 = 97u8;
pub const PANIC_CHAR_16: u16 = 44032; // u16 because its counterpart is char16_t
pub const PANIC_CHAR_32: u32 = 44032; // u32 similarly

[paxbun]

We can prevent the error by modifying <binden::ir::var::Var as bindgen::parse::ClangSubItemParser>::parse in var.rs as follows:

let (type_kind, val) = match value {
    ...
    EvalResult::Char(c) => {
        let c = match c {
            CChar::Char(c) => c as u64,
            CChar::Raw(c) => c,
        };

        if c <= ::std::u8::MAX as u64 {
            (TypeKind::Int(IntKind::U8), VarType::Char(c as u8))
        } else {
            assert!(c <= ::std::u32::MAX as u64);
            (TypeKind::Int(IntKind::U32), VarType::Int(c as i64))
        }
    }
    ...
}

But this makes PANIC_CHAR_16 and PANIC_CHAR_32 all u32.

[paxbun]

It seems like I have to modify cexpr to achieve my objectives, as CChar lacks enum variants to distinguish different character literals. If it seems okay, I would like to make a PR that adds variants for char8_t, char16_t, char32_t to cexpr::literals::CChar (which will possibly benefit jethrogb/rust-cexpr#25 as well), and then make a PR for this issue.

[pvdrz]

I suspect this could be fixed by #2369

[paxbun]

I ran the modified version, but #2369 ignores PANIC_CHAR_16 and PANIC_CHAR_32. (Ignoring those macros is enough for me, though.) If #2369 is to be merged, then maybe I should make a PR for this to reitermarkus/cmacro-rs.

[paxbun]

I posted an issue for this: reitermarkus/cmacro-rs#3