ROB: Tolerate non-array outline color when reading and cloning (#3870)

Author: metsw24-max

pypdf/_doc_common.py

@@ -1050,7 +1050,15 @@ def _build_outline_item(self, node: DictionaryObject) -> Optional[Destination]:
         if outline_item:
             if "/C" in node:
                 # Color of outline item font in (R, G, B) with values ranging 0.0-1.0
-                outline_item[NameObject("/C")] = ArrayObject(FloatObject(c) for c in node["/C"])  # type: ignore[attr-defined]
+                color = node["/C"]
+                if isinstance(color, list):
+                    outline_item[NameObject("/C")] = ArrayObject(FloatObject(c) for c in color)
+                else:
+                    logger_warning(
+                        "Ignoring non-array outline color %(color)r",
+                        source=__name__,
+                        color=color,
+                    )
             if "/F" in node:
                 # specifies style characteristics bold and/or italic
                 # with 1=italic, 2=bold, 3=both

pypdf/_writer.py

@@ -3139,11 +3139,10 @@ def _clone_outline(self, dest: Destination) -> TreeObject:
         # TODO: /SE
         if dest.node is not None:
             n_ol[NameObject("/F")] = NumberObject(dest.node.get("/F", 0))
-            n_ol[NameObject("/C")] = ArrayObject(
-                dest.node.get(
-                    "/C", [FloatObject(0.0), FloatObject(0.0), FloatObject(0.0)]
-                )
-            )
+            color = dest.node.get("/C", NullObject()).get_object()
+            if not isinstance(color, list):
+                color = [FloatObject(0.0), FloatObject(0.0), FloatObject(0.0)]
+            n_ol[NameObject("/C")] = ArrayObject(color)
         return n_ol
 
     def _insert_filtered_outline(

tests/test_doc_common.py

@@ -694,3 +694,38 @@ def test_get_qualified_field_name__cyclic() -> None:
             match=r"^Detected cycle in /Parent hierarchy when retrieving qualified field name\.$"
     ):
         _ = writer._get_qualified_field_name(parent=dictionary1)
+
+
+def test_build_outline_item__non_array_color(caplog):
+    """A malformed outline item whose /C is not an array must not crash."""
+    writer = PdfWriter()
+    writer.add_blank_page(72, 72)
+
+    item = DictionaryObject()
+    writer._add_object(item)
+    item.update({
+        NameObject("/Title"): TextStringObject("Bad color"),
+        NameObject("/Dest"): ArrayObject(
+            [writer.pages[0].indirect_reference, NameObject("/Fit")]
+        ),
+        NameObject("/C"): NumberObject(5),  # should be an [r, g, b] array
+    })
+    outlines = DictionaryObject()
+    writer._add_object(outlines)
+    outlines.update({
+        NameObject("/Type"): NameObject("/Outlines"),
+        NameObject("/First"): item.indirect_reference,
+        NameObject("/Last"): item.indirect_reference,
+    })
+    item[NameObject("/Parent")] = outlines.indirect_reference
+    writer._root_object[NameObject("/Outlines")] = outlines.indirect_reference
+
+    outline = writer.outline
+    assert outline[0]["/Title"] == "Bad color"
+    assert "/C" not in outline[0]
+    assert any("outline color" in message for message in caplog.messages)
+
+    # The same value is read again, unsanitised, while cloning the outline
+    # during append, so that path must tolerate it too.
+    target = PdfWriter()
+    target.append(writer)