<regex>: Allow initial ] to start character ranges in POSIX regular expressions (#5364)

muellerj2 · StephanTLavavej · web-flow · commit 5a1cbde572e0 · 2025-04-09T20:56:22.000-07:00
Co-authored-by: Stephan T. Lavavej &lt;stl@microsoft.com&gt;
diff --git a/stl/inc/regex b/stl/inc/regex
@@ -1734,7 +1734,7 @@ private:
     void _Do_ex_class(_Meta_type);
     bool _CharacterClassEscape(bool);
     _Prs_ret _ClassEscape2();
-    _Prs_ret _ClassAtom();
+    _Prs_ret _ClassAtom(bool);
     void _ClassRanges();
     void _CharacterClass();
     bool _IdentityEscape();
@@ -4129,7 +4129,7 @@ _Prs_ret _Parser<_FwdIt, _Elem, _RxTraits>::_ClassEscape2() { // check for class
 }
 
 template <class _FwdIt, class _Elem, class _RxTraits>
-_Prs_ret _Parser<_FwdIt, _Elem, _RxTraits>::_ClassAtom() { // check for class atom
+_Prs_ret _Parser<_FwdIt, _Elem, _RxTraits>::_ClassAtom(const bool _Initial) { // check for class atom
     if (_Mchar == _Meta_esc) { // check for valid escape sequence
         _Next();
         if (_L_flags & _L_grp_esc) {
@@ -4152,7 +4152,10 @@ _Prs_ret _Parser<_FwdIt, _Elem, _RxTraits>::_ClassAtom() { // check for class at
             _Val = _Meta_lsq;
             return _Prs_chr;
         }
-    } else if (_Mchar == _Meta_rsq || _Mchar == _Meta_eos) {
+    } else if ((_Mchar == _Meta_rsq
+                   && (!(_L_flags & _L_brk_rstr)
+                       || !_Initial)) // initial ] does not close the class when it is not special
+               || _Mchar == _Meta_eos) {
         return _Prs_none;
     } else { // handle ordinary character
         _Val = _Char;
@@ -4165,10 +4168,12 @@ template <class _FwdIt, class _Elem, class _RxTraits>
 void _Parser<_FwdIt, _Elem, _RxTraits>::_ClassRanges() { // check for valid class ranges
     _Prs_ret _Ret;
 
+    bool _Initial = true;
     for (;;) { // process characters through end of bracket expression
-        if ((_Ret = _ClassAtom()) == _Prs_none) {
+        if ((_Ret = _ClassAtom(_Initial)) == _Prs_none) {
             return;
         }
+        _Initial = false;
 
         if (_Ret == _Prs_chr && _Val == 0 && !(_L_flags & _L_bzr_chr)) {
             _Error(regex_constants::error_escape);
@@ -4178,7 +4183,7 @@ void _Parser<_FwdIt, _Elem, _RxTraits>::_ClassRanges() { // check for valid clas
             _Next();
             _Elem _Chr1               = static_cast<_Elem>(_Val);
             const bool _Set_preceding = _Ret == _Prs_set;
-            if ((_Ret = _ClassAtom()) == _Prs_none) { // treat - as ordinary character
+            if ((_Ret = _ClassAtom(false)) == _Prs_none) { // treat - as ordinary character
                 if (!_Set_preceding) {
                     _Nfa._Add_char_to_class(_Chr1);
                 }
@@ -4227,10 +4232,6 @@ void _Parser<_FwdIt, _Elem, _RxTraits>::_CharacterClass() { // add bracket expre
         _Next();
     }
 
-    if ((_L_flags & _L_brk_rstr) && _Mchar == _Meta_rsq) { // insert initial ] when not special
-        _Nfa._Add_char_to_class(_Meta_rsq);
-        _Next();
-    }
     _ClassRanges();
 }
 
diff --git a/tests/std/tests/VSO_0000000_regex_use/test.cpp b/tests/std/tests/VSO_0000000_regex_use/test.cpp
@@ -1360,6 +1360,46 @@ void test_gh_5362() {
     test_gh_5362_grep();
 }
 
+void test_gh_5364() {
+    // GH-5364 <regex>: Allow initial ] to start character ranges in POSIX regular expressions
+    for (syntax_option_type option : {basic, extended, awk, grep, egrep}) {
+        g_regexTester.should_match("]", "[]-_]", option);
+        g_regexTester.should_match("^", "[]-_]", option);
+        g_regexTester.should_match("_", "[]-_]", option);
+        g_regexTester.should_not_match("-", "[]-_]", option);
+
+        g_regexTester.should_not_match("]", "[^]-_]", option);
+        g_regexTester.should_not_match("^", "[^]-_]", option);
+        g_regexTester.should_not_match("_", "[^]-_]", option);
+        g_regexTester.should_match("-", "[^]-_]", option);
+
+        g_regexTester.should_match("]", "[]a]", option);
+        g_regexTester.should_match("a", "[]a]", option);
+        g_regexTester.should_not_match("_", "[]a]", option);
+        g_regexTester.should_not_match("a]", "[]a]", option);
+        g_regexTester.should_not_match("]a", "[]a]", option);
+        g_regexTester.should_not_match("__", "[]a]", option);
+
+        g_regexTester.should_not_match("]", "[^]a]", option);
+        g_regexTester.should_not_match("a", "[^]a]", option);
+        g_regexTester.should_match("_", "[^]a]", option);
+        g_regexTester.should_not_match("a]", "[^]a]", option);
+        g_regexTester.should_not_match("]a", "[^]a]", option);
+        g_regexTester.should_not_match("__", "[^]a]", option);
+
+        g_regexTester.should_throw("[]", error_brack, option);
+        g_regexTester.should_throw("[^]", error_brack, option);
+    }
+
+    g_regexTester.should_throw("[]-_]", error_brack, ECMAScript);
+    g_regexTester.should_throw("[^]-_]", error_brack, ECMAScript);
+    g_regexTester.should_throw("[]a]", error_brack, ECMAScript);
+    g_regexTester.should_throw("[^]a]", error_brack, ECMAScript);
+
+    g_regexTester.should_not_match("c", "[]", ECMAScript);
+    g_regexTester.should_match("c", "[^]", ECMAScript);
+}
+
 int main() {
     test_dev10_449367_case_insensitivity_should_work();
     test_dev11_462743_regex_collate_should_not_disable_regex_icase();
@@ -1398,6 +1438,7 @@ int main() {
     test_gh_5214();
     test_gh_5253();
     test_gh_5362();
+    test_gh_5364();
 
     return g_regexTester.result();
 }
