Track instruction verification bounds only for matched methods

src/asc_core/findrefs/locator/insn_locator.py

@@ -175,7 +175,7 @@ class InsnLocator(BaseLocator):
         insn_maps = self.insn_maps
 
         buf = self.buf
-        method_bounds = self.method_bounds.copy()
+        method_bounds = {}
         for off in offsets:
             midx = insn_maps.get(off >> 4)
             """
@@ -190,10 +190,10 @@ class InsnLocator(BaseLocator):
             # which means there is no possible to backtracking inside one method, so we can update method
             # start insn offset once we fullmatch an insn, avoid re-fullmatch from start of method
             # bugfix for insn mismatch issue 20260729
-            if isinstance(midx, int) and InsnLocator.INSN_VERIFY.fullmatch(buf, method_bounds[midx], off):
+            if isinstance(midx, int) and InsnLocator.INSN_VERIFY.fullmatch(buf, method_bounds.get(midx, self.method_bounds[midx]), off):
                 ret_table.append(midx)
                 method_bounds[midx] = off
-            elif isinstance(midx, list) and InsnLocator.INSN_VERIFY.fullmatch(buf, method_bounds[midx[0]], off):
+            elif isinstance(midx, list) and InsnLocator.INSN_VERIFY.fullmatch(buf, method_bounds.get(midx[0], self.method_bounds[midx[0]]), off):
                 ret_table.append(midx)
                 method_bounds[midx[0]] = off
             else:

tests/test_references.py

@@ -24,6 +24,17 @@ class ReferenceTests(unittest.TestCase):
         self.assertEqual(len(results), 1)
         self.assertEqual(set(results[0]), {0, 1})
 
+    def test_repeated_scans_preserve_method_bounds(self):
+        locator = InsnLocator(DEX.parse(memoryview(make_dex()), 'fixture.dex'))
+        locator.parse()
+        original_bounds = dict(locator.method_bounds)
+        scanner = CodeItemScanner(locator)
+        for _ in range(2):
+            query = {'string': {5}}
+            scanner.scan(query)
+            self.assertEqual(set(query['string'][0]), {0, 1})
+            self.assertEqual(locator.method_bounds, original_bounds)
+
     def test_fallback_without_class_data_map_entry(self):
         data = bytearray(make_dex())
         map_off = struct.unpack_from('<I', data, 52)[0]