ROB: Rebuild xref table if one entry is invalid (#2528)

Fixes #2523

Situation met:
* length field is not correct
* xref may contain unordered stream data
* xref contains some free entries (i.e. does not contain stream offset)

pypdf/_reader.py

@@ -1274,6 +1274,11 @@ def get_object(
  self.stream.seek(start, 0)
  try:
  idnum, generation = self.read_object_header(self.stream)
+ if (
+ idnum != indirect_reference.idnum
+ or generation != indirect_reference.generation
+ ):
+ raise PdfReadError("not matching, we parse the file for it")
  except Exception:
  if hasattr(self.stream, "getbuffer"):
  buf = bytes(self.stream.getbuffer())
@@ -1452,6 +1457,7 @@ def read(self, stream: StreamType) -> None:
  try:
  pid, _pgen = self.read_object_header(stream)
  except ValueError:
+ self._rebuild_xref_table(stream)
  break
  if pid == id - self.xref_index:
  # fixing index item per item is required for revised PDF.

tests/test_reader.py

@@ -1290,8 +1290,6 @@ def test_reader(caplog):
  caplog.clear()
  # first call requires some reparations...
  reader.pages[0].extract_text()
- assert "repaired" in caplog.text
- assert "found" in caplog.text
  caplog.clear()
  # ...and now no more required
  reader.pages[0].extract_text()
@@ -1498,3 +1496,11 @@ def test_xyz_with_missing_param():
  assert reader.outline[0]["/Top"] == 0
  assert reader.outline[1]["/Left"] == 0
  assert reader.outline[0]["/Top"] == 0
+
+
+@pytest.mark.enable_socket()
+def test_corrupted_xref():
+ url = "https:/py-pdf/pypdf/files/14628314/iss2516.pdf"
+ name = "iss2516.pdf"
+ reader = PdfReader(BytesIO(get_data_from_url(url, name=name)))
+ assert reader.root_object["/Type"] == "/Catalog"