Fix error reporting for multibyte characters in byte string literal

[tapanprakasht]

Fixes #139226

• Removed the Unicode code point suggestion as it was providing incorrect results for multibyte Unicode characters.
• The previous implementation was inconsistent—for example, it suggested a unicode code point for "µ" but gave a different suggestion for the character 字.

[rustbot]

r? @Nadrieril

rustbot has assigned @Nadrieril.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[xizheyin]

I think there may also need to be consideration of the situation in the origin code, where it's possible that users may want to use single bytes, so we can explicitly state that the suggested value is a single byte value, not a UTF-8 encoding.
So, we could add this suggestion above if mode == Mode::Byte { such as below.

          if (c as u32) <= 0xFF && mode != Mode::RawByteStr {
              err.span_suggestion(
                  span,
                  format!(
                      "if you meant to use the byte with hex value {:#04X} (note: this is not the Unicode code point for {c:?})", 
                      c as u32
                  ),
                  format!("\\x{:02X}", c as u32),
                  Applicability::MaybeIncorrect,
              );
          }

[tapanprakasht]

I believe that if a user explicitly specifies a character using a byte literal, their intention is likely to use the actual byte encoding. Otherwise, they would have directly provided the correct hex value.
Additionally, when dealing with a single byte value, it suggests a Unicode code point in hex even though that's not the full encoding, while for a multi-byte value, it throws a different error. That's more of an inconsistent behaviour.

[Nadrieril]

I don't know enough about typical uses of byte chars. We should at the very least explicitly mention that what we're suggesting is not an utf8 encoding? For reference, this behavior was introduced in #87659 to fix #87397.

[Nadrieril]

Is it true that we can only reach this branch for multibyte characters? My limited knowledge of utf8 would suggest we do the obviously-correct thing:

if mode != Mode::RawByteStr {
	let mut utf8 = String::new();
	utf8.push(c);
	if mode == Mode::Byte && utf8.as_bytes().count() > 1 {
	    // error "this is multibyte"
	} else {
	    // suggest \xHH
	}
}

[Nadrieril]

Cf my comments I'm not yet confident this is implemented correctly, but I'm ok with the idea that we should only suggest utf8-encoded things for the sake of consistency.

cc @davidtwco because you merged #87659 that introduced the previous behavior.

[Nadrieril]

@rustbot author

[rustbot]

Reminder, once the PR becomes ready for a review, use @rustbot ready.

[alex-semenyuk]

@tapanprakasht
Thanks for your contribution
From wg-triage. Any updates on this PR?