ROB: Repair PDF with invalid Root object (#2880)

Closes #2875.

Author: pubpub-zz

pypdf/_doc_common.py

@@ -1148,8 +1148,9 @@ def _flatten(
             # Fix issue 327: set flattened_pages attribute only for
             # decrypted file
             catalog = self.root_object
-            pages = catalog["/Pages"].get_object()  # type: ignore
-            assert isinstance(pages, DictionaryObject)
+            pages = catalog.get("/Pages").get_object()  # type: ignore
+            if not isinstance(pages, DictionaryObject):
+                raise PdfReadError("Invalid object in /Pages")
             self.flattened_pages = []
 
         if PA.TYPE in pages:

pypdf/_reader.py

@@ -127,6 +127,8 @@ def __init__(
         # map page indirect_reference number to page number
         self._page_id2num: Optional[Dict[Any, Any]] = None
 
+        self._validated_root: Optional[DictionaryObject] = None
+
         self._initialize_stream(stream)
 
         self._override_encryption = False
@@ -197,10 +199,35 @@ def close(self) -> None:
     @property
     def root_object(self) -> DictionaryObject:
         """Provide access to "/Root". Standardized with PdfWriter."""
-        root = self.trailer[TK.ROOT]
-        if root is None:
-            raise PdfReadError('Cannot find "/Root" key in trailer')
-        return cast(DictionaryObject, root.get_object())
+        if self._validated_root:
+            return self._validated_root
+        root = self.trailer.get(TK.ROOT)
+        if is_null_or_none(root):
+            logger_warning('Cannot find "/Root" key in trailer', __name__)
+        elif (
+            cast(DictionaryObject, cast(PdfObject, root).get_object()).get("/Type")
+            == "/Catalog"
+        ):
+            self._validated_root = cast(
+                DictionaryObject, cast(PdfObject, root).get_object()
+            )
+        else:
+            logger_warning("Invalid Root object in trailer", __name__)
+        if self._validated_root is None:
+            logger_warning('Searching object with "/Catalog" key', __name__)
+            nb = cast(int, self.trailer.get("/Size", 0))
+            for i in range(nb):
+                try:
+                    o = self.get_object(i + 1)
+                except Exception:  # to be sure to capture all errors
+                    o = None
+                if isinstance(o, DictionaryObject) and o.get("/Type") == "/Catalog":
+                    self._validated_root = o
+                    logger_warning(f"Root found at {o.indirect_reference!r}", __name__)
+                    break
+            if self._validated_root is None:
+                raise PdfReadError("Cannot find Root object in pdf")
+        return self._validated_root
 
     @property
     def _info(self) -> Optional[DictionaryObject]:
@@ -215,11 +242,11 @@ def _info(self) -> Optional[DictionaryObject]:
             return None
         else:
             info = info.get_object()
-            if info == None:  # noqa: E711
+            if not isinstance(info, DictionaryObject):
                 raise PdfReadError(
                     "Trailer not found or does not point to document information directory"
                 )
-            return cast(DictionaryObject, info)
+            return info
 
     @property
     def _ID(self) -> Optional[ArrayObject]:

pypdf/generic/_base.py

@@ -879,5 +879,6 @@ def is_null_or_none(x: Any) -> TypeGuard[Union[None, NullObject, IndirectObject]
         True if x is None or NullObject.
     """
     return x is None or (
-        isinstance(x, PdfObject) and isinstance(x.get_object(), NullObject)
+        isinstance(x, PdfObject)
+        and (x.get_object() is None or isinstance(x.get_object(), NullObject))
     )

tests/test_reader.py

@@ -137,14 +137,14 @@ def test_iss1943():
 def test_broken_meta_data(pdf_path):
     with open(pdf_path, "rb") as f:
         reader = PdfReader(f)
-        with pytest.raises(
-            PdfReadError,
-            match=(
-                "Trailer not found or does not point to document "
-                "information directory"
-            ),
-        ):
-            reader.metadata
+        assert reader.metadata is None
+
+    with open(RESOURCE_ROOT / "crazyones.pdf", "rb") as f:
+        b = f.read(-1)
+    reader = PdfReader(BytesIO(b.replace(b"/Info 2 0 R", b"/Info 2    ")))
+    with pytest.raises(PdfReadError) as exc:
+        reader.metadata
+    assert "does not point to document information directory" in repr(exc)
 
 
 @pytest.mark.parametrize(
@@ -621,7 +621,7 @@ def test_read_unknown_zero_pages(caplog):
     assert normalize_warnings(caplog.text) == warnings
     with pytest.raises(PdfReadError) as exc:
         len(reader.pages)
-    assert exc.value.args[0] == 'Cannot find "/Root" key in trailer'
+    assert exc.value.args[0] == "Invalid object in /Pages"
 
 
 def test_read_encrypted_without_decryption():
@@ -1712,3 +1712,67 @@ def test_unbalanced_brackets_in_dictionary_object(caplog):
     name = "iss2877.pdf"  # reused
     reader = PdfReader(BytesIO(get_data_from_url(url, name=name)))
     assert len(reader.pages) == 43  # note:  /Count = 46 but 3 kids are None
+
+
+@pytest.mark.enable_socket()
+def test_repair_root(caplog):
+    """Cf #2877"""
+    url = "https://github.com/user-attachments/files/17162216/crash-6620e8b1abfe3da639b654595da859b87f985748.pdf"
+    name = "iss2875.pdf"
+
+    b = get_data_from_url(url, name=name)
+    reader = PdfReader(BytesIO(b))
+    assert len(reader.pages) == 1
+    assert all(
+        msg in caplog.text
+        for msg in (
+            "Invalid Root object",
+            'Searching object with "/Catalog" key',
+            "Root found at IndirectObject(2, 0,",
+        )
+    )
+
+    # no /Root Entry
+    reader = PdfReader(BytesIO(b.replace(b"/Root", b"/Roo ")))
+    caplog.clear()
+    assert len(reader.pages) == 1
+    assert all(
+        msg in caplog.text
+        for msg in (
+            'Cannot find "/Root" key in trailer',
+            'Searching object with "/Catalog" key',
+            "Root found at IndirectObject(2, 0,",
+        )
+    )
+
+    # Invalid /Root Entry
+    caplog.clear()
+    reader = PdfReader(
+        BytesIO(
+            b.replace(b"/Root 1 0 R", b"/Root 2 0 R").replace(b"/Catalog", b"/Catalo ")
+        )
+    )
+    with pytest.raises(PdfReadError):
+        len(reader.pages)
+    assert all(
+        msg in caplog.text
+        for msg in (
+            "Invalid Root object in trailer",
+            'Searching object with "/Catalog" key',
+        )
+    )
+
+    # Invalid /Root Entry + error in get_object
+    caplog.clear()
+    b = b.replace(b"/Root 1 0 R", b"/Root 2 0 R").replace(b"/Catalog", b"/Catalo ")
+    b = b[:5124] + b"A" + b[5125:]
+    reader = PdfReader(BytesIO(b))
+    with pytest.raises(PdfReadError):
+        len(reader.pages)
+    assert all(
+        msg in caplog.text
+        for msg in (
+            "Invalid Root object in trailer",
+            'Searching object with "/Catalog" key',
+        )
+    )