v0.10.0

v0.10.0 - 2026-10-27

Added

• Problem matchers read their regular expressions as GitHub's runner does (.NET, with RegexOptions.CultureInvariant | ECMAScript), so patterns that were refused or misread now work: lookahead and lookbehind, backreferences (numbered and named), named groups numbered as .NET numbers them, \b, atomic groups, conditionals, inline options ((?i) and the rest), class subtraction, octal escapes (\033, which colour-coded linter matchers use, never matched before), \x, \u, \c, and matching by character rather than by byte (a class like [→↗] or . against é). All 77 distinct patterns in 80 problem matcher files published on GitHub, and about 140000 random patterns' inputs, answer as .NET does. \p{…} (Unicode categories) is still refused, loudly.