Use non-local OS error strings (en-US)

[liigo]

Closes #34318

[rust-highfive]

r? @aturon

(rust_highfive has picked a reviewer for you, use r? to override)

[tbu-]

Is there precedence of other programming languages doing this? From the top of my head, I only know that Python doesn't.

I don't think this is a good idea, we shouldn't impose English upon the user. If you have, say, a Windows in German, then it is expected that all strings by the operating system actually are in German.

[liigo]

@tbu- Yes, I'm Chinese, I would prefer reading OS error strings in Chinese. But Rust doesn't really print OS error strings in Chinese, it print them as \u{xxxx}\u{yyyy}..., which I can't read. English is better than \u{xxxx}\u{yyyy}..., at least. See #34318.

[tbu-]

Maybe we should rather escape less characters in the Debug implementation of &str.

[retep998]

I'd rather disable escaping only for OS error strings specifically, not for all strings.

[tbu-]

@retep998 I believe Python has a good strategy here, and they don't escape most Unicode symbols, except for weird ones like zero-width space (U+200B):

Python 3.5.1 (default, <timestamp>) 
[GCC 5.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> "ä", "\u00e4" , "\u200b"
('ä', 'ä', '\u200b')

Equivalent Rust:

fn main() {
    println!("{:?} {:?}", "ä", "\u{200b}");
}

Outputs:

"\u{e4}" "\u{200b}"

[alexcrichton]

Thanks for the PR @liigo! The libs team got a chance to talk about this today and the conclusion was that we're going to go with a solution like #34485 instead to keep localized error strings but try to escape fewer characters.