mirror of
https://github.com/LadybirdBrowser/ladybird.git
synced 2026-04-19 10:20:22 +00:00
Unicode-set intersection and subtraction always lowered their
post-consumption checks as lookbehinds. That is correct while the
outer matcher runs forward, but inside lookbehind the consumed text
sits to the right of the current position, so the checks must flip
to lookahead instead. Because we always looked left, patterns like
`(?<=[[^A-Z]--[A-Z]])P{N}` and the reported fuzz case missed
matches whenever the character before the consumed one changed the
set-operation result.
Preserve the surrounding match direction when compiling those
checks, and add coverage for reduced subtraction and intersection
cases plus the original regression.
|
||
|---|---|---|
| .. | ||
| src | ||
| build.rs | ||
| Cargo.toml | ||
| cbindgen.toml | ||