Warn about strings that skip multiple newlines with one backslash

[joshtriplett]

(This came up in today's @rust-lang/lang meeting.)

Given the following code:

fn main() {
    let s1 = "a\

b";
    let s2 = "a\

    b";
    println!("{:?} {:/}", s1, s2);
}

Playground link

The compiler interprets both of the string literals as "ab", omitting multiple newlines after the \.

This seems confusing, and different from the behavior of other languages. For instance, C and Python both produce parse errors for such string literals. Or, for a Python multi-line string literal, the backslash escapes the immediately following newline but not the subsequent newline.

Rust does allow unescaped newlines in string literals, which is useful. The ability to escape a newline can also be useful sometimes. But the combination of both escaped and unescaped newlines seems sufficiently confusing and surprising to merit a warning.

We'd ideally like to work towards this being an error. Depending on the presence of existing usage in the ecosystem, that might or might not require an edition.

[pierwill]

@rustbot claim

[pierwill]

Discussed in Zulip:

The escape sequence handling for string literals is handled here:

rust/compiler/rustc_lexer/src/unescape.rs

Line 267 in 1158367

fn unescape_str_or_byte_str<F>(src: &str, mode: Mode, callback: &mut F)

It looks like the issue lies in

rust/compiler/rustc_lexer/src/unescape.rs

Lines 300 to 307 in 1158367

	fn skip_ascii_whitespace(chars: &mut Chars<'_>) {
	let str = chars.as_str();
	let first_non_space = str
	.bytes()
	.position(|b| b != b' ' && b != b'\t' && b != b'\n' && b != b'\r')
	.unwrap_or(str.len());
	*chars = str[first_non_space..].chars()
	}

- we consider \n to be 'ascii whitespace' and will therefore skip over it after we encounter an escaped newline