1
Fork 0
mirror of https://github.com/RGBCube/serenity synced 2025-07-25 03:57:43 +00:00

LibTextCodec/Latin1: Iterate over input string with u8 instead of char

Using char causes bytes equal to or over 0x80 to be treated as a
negative value and produce incorrect results when implicitly casting to
u32.

For example, `atob` in LibWeb uses this decoder to convert non-ASCII
values to UTF-8, but non-ASCII values are >= 0x80 and thus produces
incorrect results in such cases:
```js
Uint8Array.from(atob("u660"), c => c.charCodeAt(0));
```
This used to produce [253, 253, 253] instead of [187, 174, 180].

Required by Cloudflare's IUAM challenges.
This commit is contained in:
Luke Wilde 2023-02-28 03:47:40 +00:00 committed by Sam Atkins
parent 1c918e826c
commit e864444fe3

View file

@ -353,7 +353,7 @@ ErrorOr<String> UTF16LEDecoder::to_utf8(StringView input)
ErrorOr<void> Latin1Decoder::process(StringView input, Function<ErrorOr<void>(u32)> on_code_point)
{
for (auto ch : input) {
for (u8 ch : input) {
// Latin1 is the same as the first 256 Unicode code_points, so no mapping is needed, just utf-8 encoding.
TRY(on_code_point(ch));
}