llvm-project

9cfb4172 - [libc][regex] Add POSIX character classes and shorthands (#234991)

[libc][regex] Add POSIX character classes and shorthands (#234991) POSIX ERE character class expressions like [:alpha:] and [:digit:] allow portable character classifications inside bracket expressions, while common escape shorthands (\d, \D, \s, \S, \w, \W) provide concise alternatives for character classes. In POSIX regular expressions, character classes inside brackets must correctly reject unknown class names with REG_ECTYPE, reject range endpoints involving classes with REG_ERANGE, and integrate with negated bracket sets and REG_NEWLINE. Add regex_charclass.h containing lookup_posix_class to map all 12 POSIX classes (alnum, alpha, blank, cntrl, digit, graph, lower, print, punct, space, upper, xdigit) to constexpr 256-bit bitsets in the POSIX/C locale. Support operator|= in bitset.h for bitset composition. In parse_bracket_expr, recognize [:name:] sequences, enforce REG_ECTYPE on invalid names, reject class endpoints with REG_ERANGE, and reserve word boundary anchors [[:<:]] and [[:>:]] with REG_BADPAT until word anchor support lands. Support shorthands \d, \D, \s, \S, \w, and \W in parse_quantified_atom, clearing newlines on negated shorthands when REG_NEWLINE is set. External compliance suites: Henry Spencer suite passes 126/456 (+10) boolean/error cases and 55/456 (+9) full-position cases; AT&T testregex passes 74/400 (+3) ERE boolean/error cases and 76/473 (+5) overall boolean/error cases with 0 regressions. Assisted-by: Automated tooling, human reviewed.