Infinite loop while reading metadata

[MartinThoma]

When I try to read the metadata of Effective Java 3rd Edition by Joshua Bloch.pdf it takes extremely long. It might even be an infinite loop.

Environment

Which environment were you using when you encountered the problem?

$ python -m platform
Linux-5.4.0-125-generic-x86_64-with-debian-bullseye-sid

$ python -c "import PyPDF2;print(PyPDF2.__version__)"
2.10.5

Code + PDF

This is a minimal, complete example that shows the issue:

from PyPDF2 import PdfReader

reader = PdfReader("Effective Java 3rd Edition by Joshua Bloch.pdf")
metadata = reader.metadata

The PDF: Effective Java 3rd Edition by Joshua Bloch.pdf

Affected

• NOT: PyPDF2<=2.10.4 (throws an exception)
• NOT PyPDF2>=2.10.6 (reads data properly)
• IS AFFECTED: PyPDF2==2.10.5
• See GHSA-hm9v-vj3r-r55m

[MartinThoma]

Interestingly, extracting the text works just fine:

text = ""
for i, page in enumerate(reader.pages, start=1):
    t0 = time.time()
    text += page.extract_text()
    t1 = time.time()
    print(f"Reading page {i}: {t1-t0:.2f} seconds")

[MartinThoma]

It's the line obj = self.trailer[TK.INFO] that takes so long

[MartinThoma]

generic._data_structures.read_object seems to be the issue:

def read_object(
    stream: StreamType,
    pdf: Any,  # PdfReader
    forced_encoding: Union[None, str, List[str], Dict[int, str]] = None,
) -> Union[PdfObject, int, str, ContentStream]:
    tok = stream.read(1)
    stream.seek(-1, 1)  # reset to start
    idx = ObjectPrefix.find(tok)
    if ....
    elif idx == 7:
        # comment
        while tok not in (b"\r", b"\n"):
            tok = stream.read(1)
            # Prevents an infinite loop by raising an error if the stream is at
            # the EOF
            if len(tok) <= 0:
                raise PdfStreamError("File ended unexpectedly.")
        tok = read_non_whitespace(stream)
        stream.seek(-1, 1)
        return read_object(stream, pdf, forced_encoding)

[MartinThoma]

@pubpub-zz In case you have some time, I would say this is a pretty important issue.

[pubpub-zz]

Got it!
the file is corrupt within the /Info leading to infinite loop. Cope with such situation
fixed also a warning on cmap encoding

some other possible issues detected and injected

My concern is about test: the test PDF file does not respect the licensing. I would not like to refer it for testing

[MartinThoma]

The fix was in PyPDF2/generic/_data_structures.py:

The elif tok in b"0123456789+-.": and

    else:
        raise PdfReadError("...")

broke the infinite loop