mruby-regexp: share one class per codepoint for /i literals - #7262

Under /i a literal with case counterparts compiles to a character class holding it and them, and every occurrence made a class of its own. A class id is a uint8_t, so a pattern holds 256 at most, and a phrase of a few hundred letters ran out of them and was refused. CRuby compiles and matches it.

Regexp.new("д" * 300, Regexp::IGNORECASE) =~ "Д" * 300
# CRuby:        0
# mruby before: RegexpError (too many character classes: /ддд.../)
# mruby after:  0
Regexp.new("a" * 300, Regexp::IGNORECASE) =~ "A" * 300
# CRuby:        0
# mruby before: RegexpError (too many character classes: /aaa.../)
# mruby after:  0

The class such a literal needs holds the codepoint and its case counterparts and nothing else: neither the flags in force nor the pattern around it reach it, and nothing writes to it once emit_char() or emit_cp_folded() has returned. What it holds is a function of the codepoint, so the second occurrence of a codepoint can name the class the first one made rather than make another.

Fix

The compiler records, by class id, the codepoint each /i literal class stands for (literal_cp[RE_MAX_CLASSES] in re_compiler), and literal_class() looks a codepoint up there before adding a class. The record sits in the compiler's frame and dies with it; zero marks a class made by anything else, which U+0000 cannot be confused with, since it has no case. Both literal paths, the ASCII letter in emit_char() and the folded codepoint in emit_cp_folded(), fill the class only when it is new. add_class() is unchanged: the slot is still zero cleared before num_classes counts it.

The cap now counts distinct codepoints. A pattern that folds more than 256 different ones is still refused, which CRuby does not do; that is out of scope here and the test pins the present meaning of the cap. Bracket classes ([k]) and the shorthands (\d) keep taking a class per occurrence, so "\\d" * 300 still hits the cap; that is a separate defect of the same shape and not touched here.

Testing

• mrbgems/mruby-regexp/test/unicode_case.rb (runs where RE_UNICODE_CASE is on): "д" * 300 under /i compiles and matches both cases; "дД" * 150, "д" * 256 + "a" * 256, and a \u{434} escape mixed with the spelled character all compile; the same codepoint outside (?i:...) still matches its own case alone; 281 distinct cased codepoints (Cyrillic, Latin-1, Greek, Armenian) are still refused, 207 of them compile and match their upcase.
• mrbgems/mruby-regexp/test/regexp_syntax.rb (every build): "a" * 300 and "aA" * 150 under /i compile and match; (?i:a)a matches "Aa" and not "AA".
• Both new tests fail on master with RegexpError: too many character classes and pass at the tip.

Full suite green at every commit (single commit).

Environment

# full-debug
gcc -std=gnu99 -g -O3 -Wall -Wundef -Werror-implicit-function-declaration -Wwrite-strings -g3 -O0 -DMRB_GC_STRESS -DMRB_USE_DEBUG_HOOK -DMRB_DEBUG -DMRB_USE_BIGINT -DMRB_USE_COMPLEX -DHAVE_MRUBY_ENCODING_GEM -DMRB_UTF8_STRING -DHAVE_MRUBY_IO_GEM -DMRB_USE_RATIONAL -DHAVE_MRUBY_REGEXP_GEM -DMRB_USE_SET -DMRB_USE_TASK_SCHEDULER src/string.c
# bintest
gcc -std=gnu99 -g -O3 -Wall -Wundef -Werror-implicit-function-declaration -Wwrite-strings -DMRB_GC_FIXED_ARENA -DMRB_USE_BIGINT -DMRB_USE_COMPLEX -DHAVE_MRUBY_ENCODING_GEM -DMRB_UTF8_STRING -DHAVE_MRUBY_IO_GEM -DMRB_USE_RATIONAL -DHAVE_MRUBY_REGEXP_GEM -DMRB_USE_SET -DMRB_USE_TASK_SCHEDULER -DMRB_USE_DEBUG_HOOK src/string.c
# cxx_abi
gcc -g -O3 -Wall -Wundef -Wwrite-strings -x c++ -std=gnu++03 -DMRB_GC_FIXED_ARENA -DMRB_USE_CXX_EXCEPTION -DMRB_USE_CXX_ABI -DMRB_USE_BIGINT -DMRB_USE_COMPLEX -DHAVE_MRUBY_ENCODING_GEM -DMRB_UTF8_STRING -DHAVE_MRUBY_IO_GEM -DMRB_USE_RATIONAL -DHAVE_MRUBY_REGEXP_GEM -DMRB_USE_SET -DMRB_USE_TASK_SCHEDULER src/string.c
# byte-string
gcc -std=gnu99 -g -O3 -Wall -Wundef -Werror-implicit-function-declaration -Wwrite-strings -DMRB_USE_BIGINT -DMRB_USE_COMPLEX -DHAVE_MRUBY_IO_GEM -DMRB_USE_RATIONAL -DHAVE_MRUBY_REGEXP_GEM -DMRB_USE_SET -DMRB_USE_TASK_SCHEDULER src/string.c
# ascii-case
gcc -std=gnu99 -g -O3 -Wall -Wundef -Werror-implicit-function-declaration -Wwrite-strings -DMRB_USE_ASCII_CASE -DMRB_USE_BIGINT -DMRB_USE_COMPLEX -DHAVE_MRUBY_ENCODING_GEM -DMRB_UTF8_STRING -DHAVE_MRUBY_IO_GEM -DMRB_USE_RATIONAL -DHAVE_MRUBY_REGEXP_GEM -DMRB_USE_SET -DMRB_USE_TASK_SCHEDULER src/string.c

Summary by CodeRabbit

• Bug Fixes
  • Improved case-insensitive regular expressions with many repeated characters.
  • Preserved matching behavior for ASCII and Unicode characters, including scoped /i patterns.
  • Added clear errors when regular expressions exceed the supported character-class limit.
• Tests
  • Expanded coverage for repeated literals, mixed Unicode spellings, case matching, and class-limit behavior.