<regex>: Fix matching of loops with context-dependent empty matches and bounded number of repetitions (#5820)

Co-authored-by: Stephan T. Lavavej <[email protected]>

Author: muellerj2

stl/inc/regex

@@ -4091,11 +4091,15 @@ bool _Matcher3<_BidIt, _Elem, _RxTraits, _It, _Alloc>::_Match_pat(_Node_base* _N
                             _Frame._Loop_idx_sav       = _Sav._Loop_idx;
                             _Frame._Loop_frame_idx_sav = _Sav._Loop_frame_idx;
                             _Sav._Loop_frame_idx       = _Frame_idx;
-                            if (_Progress) {
-                                ++_Sav._Loop_idx;
-                            } else { // try only one more match after an empty match
+
+                            // optimization: try only one more match after an empty match
+                            // if there is no maximum number of reps
+                            if (!_Progress && _Nr->_Max < 0) {
                                 _Sav._Loop_idx = _Nr->_Min;
+                            } else {
+                                ++_Sav._Loop_idx;
                             }
+
                             _STD fill(_Tgt_state._Grp_valid.begin() + static_cast<ptrdiff_t>(_Sav._Group_first),
                                 _Tgt_state._Grp_valid.end(), false);
                             _Next = _Nr->_Next;

tests/std/tests/VSO_0000000_regex_use/test.cpp

@@ -2260,6 +2260,13 @@ void test_gh_5792() {
     g_regexTester.should_match("bc", "(?:(?!ab))+bc");
 }
 
+void test_gh_5797() {
+    // GH-5797: <regex>: Loops with bounded number of repetitions and context-dependent empty alternative are mishandled
+    g_regexTester.should_match("bc", "(?:b|c|(?=bc)){3}");
+    g_regexTester.should_match("bc", "(^|b|c){3}");
+    g_regexTester.should_match("bc", "(^|b|c){3}", regex_constants::extended);
+}
+
 void test_gh_5798() {
     // GH-5798: <regex>: Process generic loops non-recursively.
     // This extends our test coverage on non-simple loops,
@@ -2398,6 +2405,7 @@ int main() {
     test_gh_5774();
     test_gh_5790();
     test_gh_5792();
+    test_gh_5797();
     test_gh_5798();
 
     return g_regexTester.result();