BUG: Avoid sharing array-based content streams between pages (#3681)

Otherwise, applying a transformation to one of the pages renders all
pages referencing the shared objects invalid due to
`PageObject.replace_contents` setting them to a `NullObject` without
being able to reliable check for further usages.

Closes #3680.

Author: stefan6419846

pypdf/generic/_data_structures.py

@@ -136,6 +136,13 @@ def clone(
                     force_duplicate,
                 )
                 arr.append(dup.indirect_reference)
+            elif isinstance(data, IndirectObject) and isinstance(resolved := data.get_object(), StreamObject):
+                dup = data._reference_clone(
+                    resolved.clone(pdf_dest, force_duplicate=True, ignore_fields=ignore_fields),
+                    pdf_dest,
+                    force_duplicate,
+                )
+                arr.append(dup.indirect_reference)
             elif hasattr(data, "clone"):
                 arr.append(data.clone(pdf_dest, force_duplicate, ignore_fields))
             else:

resources/Seige_of_Vicksburg_Sample_OCR-crazyones-merged.pdf

@@ -10,7 +10,16 @@
 
 from pypdf import PdfReader, PdfWriter
 from pypdf.errors import LimitReachedError
-from pypdf.generic import DictionaryObject, NameObject, RectangleObject, TreeObject
+from pypdf.generic import (
+    ArrayObject,
+    ContentStream,
+    DictionaryObject,
+    NameObject,
+    NullObject,
+    RectangleObject,
+    StreamObject,
+    TreeObject,
+)
 from tests import RESOURCE_ROOT, get_data_from_url
 
 try:
@@ -67,6 +76,50 @@ def test_array_object__clone_same_object_multiple_times(caplog):
     assert caplog.messages == []
 
 
+def test_array_object__clone_same_stream_multiple_times():
+    writer = PdfWriter()
+
+    # Unique streams.
+    stream1 = StreamObject()
+    stream1.set_data(b"Hello World!")
+    stream2 = StreamObject()
+    stream2.set_data(b"Lorem ipsum!")
+
+    # Shared streams.
+    shared_streams = [StreamObject() for _ in range(3)]
+    [shared_stream.set_data(f"Shared stream {index}".encode()) for index, shared_stream in enumerate(shared_streams)]
+
+    # Add to writer.
+    writer._add_object(stream1)
+    writer._add_object(stream2)
+    shared_references = [writer._add_object(shared_stream) for shared_stream in shared_streams]
+
+    # Arrays.
+    array1 = ArrayObject([stream1.indirect_reference, *shared_references])
+    array2 = ArrayObject([stream2.indirect_reference, *shared_references])
+
+    # Cloned.
+    cloned1 = array1.clone(pdf_dest=writer)
+    cloned2 = array2.clone(pdf_dest=writer)
+
+    # Nullify one shared object.
+    writer._replace_object(shared_references[1].indirect_reference, NullObject())
+
+    # The first entry is always different. The remaining shared entries should be dedicated copies.
+    assert cloned1[1:] != cloned2[1:]
+
+    assert ContentStream(stream=array1, pdf=None).get_data() == b"Hello World!\nShared stream 0\nShared stream 2\n"
+    assert ContentStream(stream=array2, pdf=None).get_data() == b"Lorem ipsum!\nShared stream 0\nShared stream 2\n"
+    assert (
+        ContentStream(stream=cloned1, pdf=None).get_data() ==
+        b"Hello World!\nShared stream 0\nShared stream 1\nShared stream 2\n"
+    )
+    assert (
+        ContentStream(stream=cloned2, pdf=None).get_data() ==
+        b"Lorem ipsum!\nShared stream 0\nShared stream 1\nShared stream 2\n"
+    )
+
+
 @pytest.mark.enable_socket
 def test_dictionary_object__read_from_stream__limit():
     name = "read_from_stream__length_2gb.pdf"

tests/generic/test_data_structures.py

@@ -1524,8 +1524,9 @@ def test_replace_contents__null_object_cloning_error():
         new_page = writer.add_page(page)
         new_page.scale_by(1)
 
-    assert isinstance(writer.get_object(50)["/Contents"], ContentStream)
-    assert isinstance(writer.get_object(51), NullObject)
+    page4_idnum = writer.pages[3].indirect_reference.idnum
+    assert isinstance(writer.get_object(page4_idnum)["/Contents"], ContentStream)
+    assert isinstance(writer.get_object(page4_idnum + 1), NullObject)
 
     data = BytesIO()
     writer.write(data)