<regex>: Do not reset matched capture groups in POSIX regexes (#5377)

muellerj2 · StephanTLavavej · web-flow · commit 6acd99035e11 · 2025-04-09T23:42:00.000-07:00
Co-authored-by: Stephan T. Lavavej &lt;stl@microsoft.com&gt;
diff --git a/stl/inc/regex b/stl/inc/regex
@@ -3641,9 +3641,13 @@ bool _Matcher<_BidIt, _Elem, _RxTraits, _It>::_Match_pat(_Node_base* _Nx) { // c
             { // record current position
                 _Node_capture* _Node                 = static_cast<_Node_capture*>(_Nx);
                 _Tgt_state._Grps[_Node->_Idx]._Begin = _Tgt_state._Cur;
-                // CodeQL [SM02323] Comparing unchanging unsigned int _Node->_Idx to decreasing size_t _Idx is safe.
-                for (size_t _Idx = _Tgt_state._Grp_valid.size(); _Node->_Idx < _Idx;) {
-                    _Tgt_state._Grp_valid[--_Idx] = false;
+                if (!(_Sflags
+                        & (regex_constants::basic | regex_constants::extended | regex_constants::grep
+                            | regex_constants::egrep | regex_constants::awk))) {
+                    // CodeQL [SM02323] Comparing unchanging unsigned int _Node->_Idx to decreasing size_t _Idx is safe.
+                    for (size_t _Idx = _Tgt_state._Grp_valid.size(); _Node->_Idx < _Idx;) {
+                        _Tgt_state._Grp_valid[--_Idx] = false;
+                    }
                 }
 
                 break;
diff --git a/tests/std/tests/VSO_0000000_regex_use/test.cpp b/tests/std/tests/VSO_0000000_regex_use/test.cpp
@@ -1457,6 +1457,47 @@ void test_gh_5374() {
     g_regexTester.should_match("bc", R"((a*)b\1c)", ECMAScript);
 }
 
+void test_gh_5377() {
+    // GH-5377 <regex>: Do not reset matched capture groups in POSIX regexes
+    for (syntax_option_type option : {extended, awk, egrep}) {
+        test_regex abcd_regex(&g_regexTester, R"(^((a)|(b)|(c)|(d))+$)", option);
+        abcd_regex.should_search_match_capture_groups(
+            "abcd", "abcd", match_default, {{3, 4}, {0, 1}, {1, 2}, {2, 3}, {3, 4}});
+        abcd_regex.should_search_match_capture_groups(
+            "acbd", "acbd", match_default, {{3, 4}, {0, 1}, {2, 3}, {1, 2}, {3, 4}});
+        abcd_regex.should_search_match_capture_groups(
+            "dcba", "dcba", match_default, {{3, 4}, {3, 4}, {2, 3}, {1, 2}, {0, 1}});
+    }
+
+    for (syntax_option_type option : {basic, grep}) {
+        test_regex abcd_regex(&g_regexTester, R"(^\(\(a\)*\(b\)*\(c\)*\(d\)*\)*$)", option);
+        abcd_regex.should_search_match_capture_groups(
+            "abcd", "abcd", match_default, {{0, 4}, {0, 1}, {1, 2}, {2, 3}, {3, 4}});
+        abcd_regex.should_search_match_capture_groups(
+            "acbd", "acbd", match_default, {{2, 4}, {0, 1}, {2, 3}, {1, 2}, {3, 4}});
+        abcd_regex.should_search_match_capture_groups(
+            "dcba", "dcba", match_default, {{3, 4}, {3, 4}, {2, 3}, {1, 2}, {0, 1}});
+
+        test_regex backref_regex(&g_regexTester, R"(^\(\(a\)\{0,1\}\(\2b\)\{0,1\}\)*)", option);
+        backref_regex.should_search_match_capture_groups("aaababb", "aaabab", match_default, {{4, 6}, {1, 2}, {4, 6}});
+    }
+
+    {
+        // ECMAScript's behavior is different:
+        test_regex abcd_regex(&g_regexTester, R"(^((a)|(b)|(c)|(d))+$)", ECMAScript);
+        abcd_regex.should_search_match_capture_groups(
+            "abcd", "abcd", match_default, {{3, 4}, {-1, -1}, {-1, -1}, {-1, -1}, {3, 4}});
+        abcd_regex.should_search_match_capture_groups(
+            "acbd", "acbd", match_default, {{3, 4}, {-1, -1}, {-1, -1}, {-1, -1}, {3, 4}});
+        abcd_regex.should_search_match_capture_groups(
+            "dcba", "dcba", match_default, {{3, 4}, {3, 4}, {-1, -1}, {-1, -1}, {-1, -1}});
+
+        test_regex backref_regex(&g_regexTester, R"(^((a){0,1}(\2b){0,1})*)", ECMAScript);
+        backref_regex.should_search_match_capture_groups(
+            "aaababb", "aaababb", match_default, {{6, 7}, {-1, -1}, {6, 7}});
+    }
+}
+
 int main() {
     test_dev10_449367_case_insensitivity_should_work();
     test_dev11_462743_regex_collate_should_not_disable_regex_icase();
@@ -1499,6 +1540,7 @@ int main() {
     test_gh_5364();
     test_gh_5371();
     test_gh_5374();
+    test_gh_5377();
 
     return g_regexTester.result();
 }
