Add test with MISSING alerts

python/ql/lib/change-notes/2026-06-11-fix-type-tracking-instance-attributes.md

@@ -1,4 +0,0 @@
----
-category: minorAnalysis
----
-* Python type tracking now follows values stored in instance attributes such as `self.attr` across instance methods on the same class. As a result, analysis is more likely to recognize user-defined objects that are stored on `self` and used later in other methods, which may produce additional results.

python/ql/lib/semmle/python/dataflow/new/internal/TypeTrackingImpl.qll

@@ -172,8 +172,6 @@ module TypeTrackingInput implements Shared::TypeTrackingInput<Location> {
   /** Holds if there is a level step from `nodeFrom` to `nodeTo`, which does not depend on the call graph. */
   predicate levelStepNoCall(Node nodeFrom, LocalSourceNode nodeTo) {
     TypeTrackerSummaryFlow::levelStepNoCall(nodeFrom, nodeTo)
-    or
-    localFieldStep(nodeFrom, nodeTo)
   }
 
   /**
@@ -319,51 +317,6 @@ module TypeTrackingInput implements Shared::TypeTrackingInput<Location> {
     )
   }
 
-  /**
-   * Holds if `ref` accesses attribute `attr` of `self`, where `self` is the first
-   * parameter of an instance method of `cls` (i.e. an access of the form `self.attr`).
-   *
-   * Static methods and class methods are excluded, since their first parameter is not a
-   * `self` instance reference.
-   */
-  private predicate selfAttrRef(Class cls, string attr, DataFlowPublic::AttrRef ref) {
-    exists(Function method, Name selfUse |
-      method = cls.getAMethod() and
-      not DataFlowDispatch::isStaticmethod(method) and
-      not DataFlowDispatch::isClassmethod(method) and
-      selfUse.getVariable() = method.getArg(0).(Name).getVariable() and
-      ref.getObject().asCfgNode().getNode() = selfUse and
-      ref.mayHaveAttributeName(attr)
-    )
-  }
-
-  /**
-   * Holds if `nodeFrom` is written to attribute `self.attr` in some instance method of a
-   * class, and `nodeTo` reads attribute `self.attr` in some (possibly different) instance
-   * method of the same class.
-   *
-   * This models flow through instance attributes (`self.foo`): a value stored into
-   * `self.foo` in one method can be read from `self.foo` in another method. Type-tracking
-   * handles the store and read steps via `AttrWrite`/`AttrRead`, but on its own it cannot
-   * relate the `self` of the writing method to the `self` of the reading method. Following
-   * the approach used for Ruby and JavaScript, we model this directly as a level step from
-   * the written value to the read reference, for any pair of methods on the class (not
-   * just from `__init__`).
-   *
-   * This is an over-approximation: it is instance-insensitive (it does not distinguish
-   * between different instances of the same class) and order-insensitive (it does not
-   * require the write to happen before the read), matching the precision of
-   * instance-attribute handling for Ruby and JavaScript.
-   */
-  private predicate localFieldStep(Node nodeFrom, LocalSourceNode nodeTo) {
-    exists(Class cls, string attr, DataFlowPublic::AttrWrite write, DataFlowPublic::AttrRead read |
-      selfAttrRef(cls, attr, write) and
-      nodeFrom = write.getValue() and
-      selfAttrRef(cls, attr, read) and
-      nodeTo = read
-    )
-  }
-
   /**
    * Holds if data can flow from `node1` to `node2` in a way that discards call contexts.
    */

python/ql/test/experimental/library-tests/CallGraph/InlineCallGraphTest.expected

@@ -1,6 +1,9 @@
 testFailures
 debug_callableNotUnique
 pointsTo_found_typeTracker_notFound
+| code/class_attr_assign.py:10:9:10:27 | ControlFlowNode for Attribute() | my_func |
+| code/class_attr_assign.py:11:9:11:25 | ControlFlowNode for Attribute() | my_func |
+| code/class_attr_assign.py:26:9:26:25 | ControlFlowNode for Attribute() | DummyObject.method |
 | code/class_super.py:50:1:50:6 | ControlFlowNode for Attribute() | outside_def |
 | code/conditional_in_argument.py:18:5:18:11 | ControlFlowNode for Attribute() | X.bar |
 | code/func_defined_outside_class.py:21:1:21:11 | ControlFlowNode for Attribute() | A.foo |

python/ql/test/experimental/library-tests/CallGraph/code/class_attr_assign.py

@@ -7,8 +7,8 @@ class Foo(object):
         self.direct_ref = my_func
 
     def later(self):
-        self.indirect_ref() # $ pt=my_func tt=my_func
+        self.indirect_ref() # $ pt=my_func MISSING: tt=my_func
-        self.direct_ref() # $ pt=my_func tt=my_func
+        self.direct_ref() # $ pt=my_func MISSING: tt=my_func
 
 foo = Foo(my_func) # $ tt=Foo.__init__
 foo.later() # $ pt,tt=Foo.later
@@ -23,7 +23,7 @@ class Bar(object):
         self.obj = DummyObject()
 
     def later(self):
-        self.obj.method() # $ pt=DummyObject.method tt=DummyObject.method
+        self.obj.method() # $ pt=DummyObject.method MISSING: tt=DummyObject.method
 
 
 bar = Bar(my_func) # $ tt=Bar.__init__

python/ql/test/library-tests/dataflow/typetracking/attribute_tests.py

@@ -151,10 +151,10 @@ class MyClass2(object):
         self.foo = tracked # $ tracked=foo tracked
 
     def print_foo(self): # $ MISSING: tracked=foo
-        print(self.foo) # $ tracked MISSING: tracked=foo
+        print(self.foo) # $ MISSING: tracked=foo tracked
 
     def possibly_uncalled_method(self): # $ MISSING: tracked=foo
-        print(self.foo) # $ tracked MISSING: tracked=foo
+        print(self.foo) # $ MISSING: tracked=foo tracked
 
 instance = MyClass2()
 print(instance.foo) # $ MISSING: tracked=foo tracked

python/ql/test/library-tests/frameworks/hdbcli/pep249.py

@@ -9,28 +9,47 @@ cursor.executemany("some sql", (42,))  # $ getSql="some sql"
 cursor.close()
 
 
-# Connection stored in a class attribute (`self._conn`) and used in another method.
+# ---------------------------------------------------------------------------
-#
+# Connection stored in a class attribute and accessed via various patterns
-# This is detected because type tracking includes a level step modelling flow through
+# ---------------------------------------------------------------------------
-# instance attributes: a value written to `self._conn` in one method (here `__init__`) can
+
-# be read back from `self._conn` (directly or via a getter) in any other method on the same
+
-# class. This follows the same approach used for instance fields in Ruby and JavaScript.
+class WrapperA:
-class Database:
     def __init__(self):
-        self._conn = dbapi.connect(address="hostname", port=300, user="username")
+        self._conn = dbapi.connect(address="hostname", port=300, user="username", pass_arg="testpass")
 
     def get_connection(self):
         return self._conn
 
-    def run_via_getter(self):
-        conn = self.get_connection()
-        cursor = conn.cursor()
-        cursor.execute("getter sql")  # $ getSql="getter sql"
 
-    def run_direct(self):
+# Getter called on a fresh constructor result
-        self._conn.execute("direct sql")  # $ getSql="direct sql"
+conn_a1 = WrapperA().get_connection()
+cursor_a1 = conn_a1.cursor()
+cursor_a1.execute("some sql", (42,))  # $ MISSING: getSql="some sql"
+
+# Getter called via a stored wrapper instance
+wrapper_instance = WrapperA()
+conn_a2 = wrapper_instance.get_connection()
+cursor_a2 = conn_a2.cursor()
+cursor_a2.execute("some sql", (42,))  # $ MISSING: getSql="some sql"
+
+# Direct attribute access on a fresh constructor result
+conn_b = WrapperA()._conn
+cursor_b = conn_b.cursor()
+cursor_b.execute("some sql", (42,))  # $ MISSING: getSql="some sql"
 
 
-db = Database()
+class WrapperB:
-db.run_via_getter()
+    """Stores the connection under a different attribute name."""
-db.run_direct()
+
+    def __init__(self):
+        self._hana = dbapi.connect(address="hostname", port=300, user="username", pass_arg="testpass")
+
+    def cursor(self):
+        return self._hana.cursor()
+
+
+# Direct attribute access on a stored instance (mirrors hdb_con3 in the issue)
+conn_c = WrapperB()._hana
+cursor_c = conn_c.cursor()
+cursor_c.execute("some sql", (42,))  # $ MISSING: getSql="some sql"