rustc_errors: enforce OUTPUT_REPLACEMENTS is sorted with a compile-time assertion

This commit is contained in:
Pavel Grigorenko 2024-08-01 00:57:38 +03:00
parent 8c7e0e1608
commit b0a8944120

View File

@ -2564,9 +2564,7 @@ fn num_decimal_digits(num: usize) -> usize {
// We replace some characters so the CLI output is always consistent and underlines aligned.
// Keep the following list in sync with `rustc_span::char_width`.
// ATTENTION: keep lexicografically sorted so that the binary search will work
const OUTPUT_REPLACEMENTS: &[(char, &str)] = &[
// tidy-alphabetical-start
// In terminals without Unicode support the following will be garbled, but in *all* terminals
// the underlying codepoint will be as well. We could gate this replacement behind a "unicode
// support" gate.
@ -2579,7 +2577,7 @@ const OUTPUT_REPLACEMENTS: &[(char, &str)] = &[
('\u{0006}', ""),
('\u{0007}', ""),
('\u{0008}', ""),
('\u{0009}', " "), // We do our own tab replacement
('\t', " "), // We do our own tab replacement
('\u{000b}', ""),
('\u{000c}', ""),
('\u{000d}', ""),
@ -2612,10 +2610,20 @@ const OUTPUT_REPLACEMENTS: &[(char, &str)] = &[
('\u{2067}', "<EFBFBD>"),
('\u{2068}', "<EFBFBD>"),
('\u{2069}', "<EFBFBD>"),
// tidy-alphabetical-end
];
fn normalize_whitespace(s: &str) -> String {
const {
let mut i = 1;
while i < OUTPUT_REPLACEMENTS.len() {
assert!(
OUTPUT_REPLACEMENTS[i - 1].0 < OUTPUT_REPLACEMENTS[i].0,
"The OUTPUT_REPLACEMENTS array must be sorted (for binary search to work) \
and must contain no duplicate entries"
);
i += 1;
}
}
// Scan the input string for a character in the ordered table above. If it's present, replace
// it with it's alternative string (it can be more than 1 char!). Otherwise, retain the input
// char. At the end, allocate all chars into a string in one operation.