[3.11] gh-100061: Proper fix of the bug in the matching of possessive quantifiers (GH-102612) (GH-108004)

Restore the global Input Stream pointer after trying to match a sub-pattern.

Co-authored-by: Ma Lin <[email protected]>

(cherry picked from commit abd9cc5)
    
Co-authored-by: SKO <[email protected]>

Author: serhiy-storchaka

Lib/re/_compiler.py

@@ -100,13 +100,6 @@ def _compile(code, pattern, flags):
                 emit(ANY_ALL)
             else:
                 emit(ANY)
-        elif op is POSSESSIVE_REPEAT:
-            # gh-106052: Possessive quantifiers do not work when the
-            # subpattern contains backtracking, i.e. "(?:ab?c)*+".
-            # Implement it as equivalent greedy qualifier in atomic group.
-            p = [(MAX_REPEAT, av)]
-            p = [(ATOMIC_GROUP, p)]
-            _compile(code, p, flags)
         elif op in REPEATING_CODES:
             if flags & SRE_FLAG_TEMPLATE:
                 raise error("internal: unsupported template operator %r" % (op,))

Lib/test/test_re.py

@@ -2397,6 +2397,16 @@ def test_template_function_and_flag_is_deprecated(self):
         self.assertFalse(template_re1.match('nope'))
 
     def test_bug_gh106052(self):
+        # gh-100061
+        self.assertEqual(re.match('(?>(?:.(?!D))+)', 'ABCDE').span(), (0, 2))
+        self.assertEqual(re.match('(?:.(?!D))++', 'ABCDE').span(), (0, 2))
+        self.assertEqual(re.match('(?>(?:.(?!D))*)', 'ABCDE').span(), (0, 2))
+        self.assertEqual(re.match('(?:.(?!D))*+', 'ABCDE').span(), (0, 2))
+        self.assertEqual(re.match('(?>(?:.(?!D))?)', 'CDE').span(), (0, 0))
+        self.assertEqual(re.match('(?:.(?!D))?+', 'CDE').span(), (0, 0))
+        self.assertEqual(re.match('(?>(?:.(?!D)){1,3})', 'ABCDE').span(), (0, 2))
+        self.assertEqual(re.match('(?:.(?!D)){1,3}+', 'ABCDE').span(), (0, 2))
+        # gh-106052
         self.assertEqual(re.match("(?>(?:ab?c)+)", "aca").span(), (0, 2))
         self.assertEqual(re.match("(?:ab?c)++", "aca").span(), (0, 2))
         self.assertEqual(re.match("(?>(?:ab?c)*)", "aca").span(), (0, 2))
@@ -2502,7 +2512,6 @@ def test_atomic_group(self):
 17: SUCCESS
 ''')
 
-    @unittest.expectedFailure  # gh-106052
     def test_possesive_repeat_one(self):
         self.assertEqual(get_debug_out(r'a?+'), '''\
 POSSESSIVE_REPEAT 0 1
@@ -2515,7 +2524,6 @@ def test_possesive_repeat_one(self):
 12: SUCCESS
 ''')
 
-    @unittest.expectedFailure  # gh-106052
     def test_possesive_repeat(self):
         self.assertEqual(get_debug_out(r'(?:ab)?+'), '''\
 POSSESSIVE_REPEAT 0 1

Misc/NEWS.d/next/Library/2023-03-14-01-19-57.gh-issue-100061.CiXJYn.rst

@@ -0,0 +1,2 @@
+Fix a bug that causes wrong matches for regular expressions with possessive
+qualifier.

Modules/_sre/sre_lib.h

@@ -1334,6 +1334,10 @@ SRE(match)(SRE_STATE* state, const SRE_CODE* pattern, int toplevel)
                     MARK_POP(ctx->lastmark);
                     LASTMARK_RESTORE();
 
+                    /* Restore the global Input Stream pointer
+                       since it can change after jumps. */
+                    state->ptr = ptr;
+
                     /* We have sufficient matches, so exit loop. */
                     break;
                 }