ROB: Cope with corrupted entries in xref table (#1300)

This robustness improvement is for PDF files that have a corrupted Xref table entry, but the object can be found in the PDF file by searching the file for the entry.

Closes #1292

PyPDF2/_reader.py

@@ -1385,10 +1385,37 @@ def _read_standard_xref_table(self, stream: StreamType) -> None:
                 if line[-1] in b"0123456789t":
                     stream.seek(-1, 1)
 
-                offset_b, generation_b = line[:16].split(b" ")
-                entry_type_b = line[17:18]
+                try:
+                    offset_b, generation_b = line[:16].split(b" ")
+                    entry_type_b = line[17:18]
+
+                    offset, generation = int(offset_b), int(generation_b)
+                except Exception:
+                    # if something wrong occured
+                    if hasattr(stream, "getbuffer"):
+                        buf = bytes(stream.getbuffer())  # type: ignore
+                    else:
+                        p = stream.tell()
+                        stream.seek(0, 0)
+                        buf = stream.read(-1)
+                        stream.seek(p)
+
+                    f = re.search(f"{num}\\s+(\\d+)\\s+obj".encode(), buf)
+                    if f is None:
+                        logger_warning(
+                            f"entry {num} in Xref table invalid; object not found",
+                            __name__,
+                        )
+                        generation = 65535
+                        offset = -1
+                    else:
+                        logger_warning(
+                            f"entry {num} in Xref table invalid but object found",
+                            __name__,
+                        )
+                        generation = int(f.group(1))
+                        offset = f.start()
 
-                offset, generation = int(offset_b), int(generation_b)
                 if generation not in self.xref:
                     self.xref[generation] = {}
                     self.xref_free_entry[generation] = {}

tests/test_reader.py

@@ -1094,3 +1094,21 @@ def test_wrong_password_error():
 def test_get_page_number_by_indirect():
     reader = PdfReader(RESOURCE_ROOT / "crazyones.pdf")
     reader._get_page_number_by_indirect(1)
+
+
+def test_corrupted_xref_table():
+    # issue #1292
+    url = "https://github.com/py-pdf/PyPDF2/files/9444747/BreezeManual.orig.pdf"
+    name = "BreezeMan1.pdf"
+    reader = PdfReader(BytesIO(get_pdf_from_url(url, name=name)))
+    reader.pages[0].extract_text()
+
+    url = "https://github.com/py-pdf/PyPDF2/files/9444748/BreezeManual.failed.pdf"
+    name = "BreezeMan2.pdf"
+    reader = PdfReader(BytesIO(get_pdf_from_url(url, name=name)))
+    try:
+        reader.pages[0].extract_text()
+    except Exception:
+        pass  # Exception normal
+    else:
+        raise Exception("page 0 should not be corrupted")