Bug 190005 – URLs with mismatched surrogate pairs in the host should fail to parse

• Component: New Bugs
• Status: RESOLVED FIXED

[Alex Christensen]

URLs with mismatched surrogate pairs in the host should fail to parse

[Alex Christensen]

Created attachment 350879 [details]
Patch

[Alex Christensen]

http://trac.webkit.org/r236528

[Radar WebKit Bug Importer]

<rdar://problem/44809426>

[Alexey Proskuryakov]

Comment on attachment 350879 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=350879&action=review

> Tools/TestWebKitAPI/Tests/WebCore/URLParser.cpp:1260
> +    const wchar_t replacementCharacter = 0xFFFD;

Can this be verified with a web facing test? API tests are a lot more costly in many ways (poor infrastructure to maintain, no EWS support, no parallelization).

[Darin Adler]

Comment on attachment 350879 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=350879&action=review

> Source/WebCore/platform/URLParser.cpp:2758
> +        if (!U_IS_UNICODE_CHAR(*iterator))

This line of code does not simply check for mismatched surrogates. If we only wanted to do that, the correct code would be:

    if (U_IS_SURROGATE(*iterator))

The U_IS_UNICODE_CHAR function excludes surrogates, but also excludes FDD0-FDEF, FFFE, and FFFF. Do we want that behavior? The added test doesn't cover this, but it should cover those cases too if we think the behavior change is helpful.