How about adding an example for the invalid Unicode case? Either way is definitely wouldn't return Cow::Borrowed it would return Cow::Owned.

I figured it'd be undefined behavior to create a string w/ invalid unicode, as your comment here.

Either way is definitely wouldn't return Cow::Borrowed it would return Cow::Owned.

Yep, good catch, I'll change that

I figured it'd be undefined behavior to create a string w/ invalid unicode

Strings and strs have to contain valid UTF-8 but CStrings and Cstrs are just sequences of any none null bytes. The following example would be fine:

use std::borrow::Cow;
use std::ffi::CStr;

let c_str = CStr::from_bytes_with_nul(b"foo\xc3\0").unwrap();
assert_eq!(c_str.to_string_lossy(), "foo�");

It would also be more correct to say "(in)valid UTF-8" rather than "(in)valid unicode" in this case.

It might be worth borrowing from the example in String::from_utf8_lossy, which uses the input string b"Hello \xF0\x90\x80World" to demonstrate the replacement.

Could you change this to ```no_run? Also the some_external_function(raw) on line 307 should be some_extern_function(raw).

I think I'll get linker errors upon building, right?

Latest force push addresses the typo, good catch

@frewsxcv Currently no_run (and compile_fail as well) will stop after analysis. It won't even run codegen.