BUG: Fix missing "PreventGC" when cloning (#3520)

Closes #3450.

---------

Co-authored-by: Thiago Bellini Ribeiro <thiago@bellini.dev>

Author: patrick91

pypdf/generic/_base.py

@@ -170,7 +170,7 @@ def _reference_clone(
         if ind is not None:
             if id(ind.pdf) not in pdf_dest._id_translated:
                 pdf_dest._id_translated[id(ind.pdf)] = {}
-                pdf_dest._id_translated[id(ind.pdf)]["PreventGC"] = ind.pdf  # type: ignore
+                pdf_dest._id_translated[id(ind.pdf)]["PreventGC"] = ind.pdf  # type: ignore[index]
             if (
                 not force_duplicate
                 and ind.idnum in pdf_dest._id_translated[id(ind.pdf)]
@@ -346,6 +346,7 @@ def clone(
             return self
         if id(self.pdf) not in pdf_dest._id_translated:
             pdf_dest._id_translated[id(self.pdf)] = {}
+            pdf_dest._id_translated[id(self.pdf)]["PreventGC"] = self.pdf  # type: ignore[index]
 
         if self.idnum in pdf_dest._id_translated[id(self.pdf)]:
             dup = pdf_dest.get_object(pdf_dest._id_translated[id(self.pdf)][self.idnum])

tests/test_generic.py

@@ -1,6 +1,8 @@
 """Test the pypdf.generic module."""
 
 import codecs
+import gc
+import weakref
 from base64 import a85encode
 from copy import deepcopy
 from io import BytesIO
@@ -916,6 +918,63 @@ def test_cloning(caplog):
     assert isinstance(obj21.get("/Test2"), IndirectObject)
 
 
+def test_cloning_indirect_obj_keeps_hard_reference():
+    """
+    Reported in #3450
+
+    Ensure that cloning an IndirectObject keeps a hard reference to
+    the underlying object, preventing its deallocation, which could allow
+    `id(obj)` to return the same value for different objects.
+    """
+    writer1 = PdfWriter()
+    indirect_object = IndirectObject(1, 0, writer1)
+
+    # Create a weak reference to the underlying object to test later
+    # if it is still alive in memory or not
+    obj_weakref = weakref.ref(indirect_object.pdf)
+    assert obj_weakref() is not None
+
+    writer2 = PdfWriter()
+    indirect_object.clone(writer2)
+
+    # Mimic indirect_object/writer1 going out of scope and being
+    # garbage collected. Clone should have kept a hard reference to
+    # it, preventing its deallocation.
+    del indirect_object
+    del writer1
+    gc.collect()
+    assert obj_weakref() is not None
+
+
+def test_cloning_null_obj_keeps_hard_reference():
+    """
+    Ensure that cloning a NullObject keeps a hard reference to
+    the underlying object, preventing its deallocation, which could allow
+    `id(obj)` to return the same value for different objects.
+    """
+    writer1 = PdfWriter()
+    indirect_object = IndirectObject(1, 0, writer1)
+    null_obj = NullObject()
+    null_obj.indirect_reference = indirect_object
+
+    # Create a weak reference to the underlying object to test later
+    # if it is still alive in memory or not
+    obj_weakref = weakref.ref(indirect_object.pdf)
+    assert obj_weakref() is not None
+
+    writer2 = PdfWriter()
+    null_obj.clone(writer2)
+
+    # Mimic indirect_object/writer1 going out of scope and being
+    # garbage collected. Clone should have kept a hard reference to
+    # it, preventing its deallocation.
+    del indirect_object
+    del writer1
+    del null_obj
+    gc.collect()
+    assert obj_weakref() is not None
+
+
 @pytest.mark.enable_socket
 def test_append_with_indirectobject_not_pointing(caplog):
     """