mirror of
https://github.com/LadybirdBrowser/ladybird
synced 2026-04-25 17:25:08 +02:00
87b22d0c04
Unicode set intersection and subtraction were compiled by matching one
operand and then checking the others with lookbehind. That let a
longer string operand reject a shorter match whenever the longer
string happened to end at the same position.
Group unicode set operands by exact match length and compile each
length class separately, longest first. This keeps longest-match
semantics for unions while making intersection and subtraction compare
only strings of the same length. The new RegExp runtime cases cover
both the reported [a-z]--\q{abc} regression and the related
intersection/subtraction mismatches, and they now agree with V8.