ROB: guard empty link destination arrays during page transfer

pypdf/_writer.py

@@ -3035,7 +3035,8 @@ def _insert_filtered_annotations(
                         if str(d) in self.get_named_dest_root():
                             outlist.append(ano.clone(self).indirect_reference)
                     else:
-                        d = cast("ArrayObject", d)
+                        if not isinstance(d, ArrayObject) or not d:
+                            continue
                         p = self._get_cloned_page(d[0], pages, reader)
                         if p is not None:
                             anc = ano.clone(self, ignore_fields=("/Dest",))
@@ -3050,7 +3051,8 @@ def _insert_filtered_annotations(
                     if str(d) in self.get_named_dest_root():
                         outlist.append(ano.clone(self).indirect_reference)
                 else:
-                    d = cast("ArrayObject", d)
+                    if not isinstance(d, ArrayObject) or not d:
+                        continue
                     p = self._get_cloned_page(d[0], pages, reader)
                     if p is not None:
                         anc = ano.clone(self, ignore_fields=("/D",))

pypdf/generic/_link.py

@@ -65,7 +65,9 @@ def __init__(self, reference: ArrayObject) -> None:
         """reference: an ArrayObject whose first element is the Page indirect object"""
         self._reference = reference
 
-    def find_referenced_page(self) -> IndirectObject:
+    def find_referenced_page(self) -> Union[IndirectObject, None]:
+        if not self._reference:
+            return None
         return cast(IndirectObject, self._reference[0])
 
     def patch_reference(self, target_pdf: "PdfWriter", new_page: IndirectObject) -> None:

tests/generic/test_link.py

@@ -120,3 +120,10 @@ def test_extract_links_ignores_uri_annotation_offsets(caplog: pytest.LogCaptureF
     assert isinstance(links[0][0], DirectReferenceLink)
     assert isinstance(links[0][1], DirectReferenceLink)
     assert caplog.messages == []
+
+
+def test_direct_reference_link_empty_reference() -> None:
+    # An empty destination array has no target page to resolve.
+    assert DirectReferenceLink(ArrayObject([])).find_referenced_page() is None
+    # A populated array still resolves to its first element.
+    assert DirectReferenceLink(ArrayObject([NumberObject(7)])).find_referenced_page() == 7

tests/test_writer.py

@@ -899,6 +899,52 @@ def test_link_annotation(pdf_file_path):
         writer.write(output_stream)
 
 
+@pytest.mark.parametrize(
+    "annotation",
+    [
+        # Empty /Dest array.
+        DictionaryObject({
+            NameObject("/Subtype"): NameObject("/Link"),
+            NameObject("/Dest"): ArrayObject([]),
+        }),
+        # Empty /GoTo action destination.
+        DictionaryObject({
+            NameObject("/Subtype"): NameObject("/Link"),
+            NameObject("/A"): DictionaryObject({
+                NameObject("/S"): NameObject("/GoTo"),
+                NameObject("/D"): ArrayObject([]),
+            }),
+        }),
+        # /Dest that is not an array at all.
+        DictionaryObject({
+            NameObject("/Subtype"): NameObject("/Link"),
+            NameObject("/Dest"): NumberObject(5),
+        }),
+    ],
+    ids=["empty-dest-array", "empty-goto-action-dest", "non-array-dest"],
+)
+def test_malformed_link_destination_does_not_crash_transfer(annotation):
+    """A link with an empty or non-array destination must not abort page transfer."""
+    writer = PdfWriter()
+    writer.add_blank_page(width=72, height=72)
+    writer.pages[0][NameObject("/Annots")] = ArrayObject([writer._add_object(annotation)])
+    stream = BytesIO()
+    writer.write(stream)
+    stream.seek(0)
+    reader = PdfReader(stream)
+
+    # append() routes the annotation through _insert_filtered_annotations.
+    appended = PdfWriter()
+    appended.append(reader)
+    appended.write(BytesIO())
+
+    # add_page() + write() routes it through extract_links()/_resolve_links().
+    added = PdfWriter()
+    for page in reader.pages:
+        added.add_page(page)
+    added.write(BytesIO())
+
+
 def test_io_streams():
     """This is the example from the docs ("Streaming data")."""
     filepath = RESOURCE_ROOT / "pdflatex-outline.pdf"