KeyError on malformed PDF

[Google-Autofuzz]

When running the following code with the latest pypi version of PyPDF2 on the attached input results in an unexpected KeyError.

edit: Updated to reflect the PyPDF2==2.4.2 version.

MCVE: Code + PDF

PDF file: test.pdf

import sys
from PyPDF2 import PdfReader

reader = PdfReader(sys.argv[1], strict=False)

reader.metadata
reader.get_fields()
reader.named_destinations
len(reader.pages)
reader.outlines
reader.page_layout
reader.page_mode
reader.xmp_metadata

Traceback

$ python pypdf2_repro.py test.pdf
Traceback (most recent call last):
  File "foo.py", line 7, in <module>
    reader.get_fields()
  File "/home/moose/Github/py-pdf/PyPDF2/PyPDF2/_reader.py", line 482, in get_fields
    catalog = cast(DictionaryObject, self.trailer[TK.ROOT])
  File "/home/moose/Github/py-pdf/PyPDF2/PyPDF2/generic.py", line 663, in __getitem__
    return dict.__getitem__(self, key).get_object()
KeyError: '/Root'

[MartinThoma]

I can confirm this wit PyPDF2==1.27.7:

Traceback (most recent call last):
  File "/home/moose/foo.py", line 8, in <module>
    pdf.getFields()
  File "/home/moose/Github/py-pdf/PyPDF2/PyPDF2/pdf.py", line 1331, in getFields
    catalog = self.trailer[TK.ROOT]
  File "/home/moose/Github/py-pdf/PyPDF2/PyPDF2/generic.py", line 539, in __getitem__
    return dict.__getitem__(self, key).getObject()
KeyError: '/Root'

According to "TABLE 3.13 Entries in the file trailer dictionary" of the PDF 1.7 Specifications, the "/Root" key is required in the trailer. So I can confirm that the PDF is malformed.

[MartinThoma]

I've just updated the example and confirmed that it still happens.

[DL6ER]

Some more examples where this is happening when accessing, e.g., pdfreader.numPages:

• qp.pdf
• convexf_def.pdf
• lr_nn.pdf
• NAG.pdf
• biastrick.pdf
• pla_nn.pdf
• pca_procedure.pdf
• softmax_nn.pdf
• pca_var0.pdf
• move2trash.pdf
• normballs.pdf
• scores.pdf
• pla_nn2.pdf
• first_order.pdf
• lda6.pdf
• planting.pdf
• max_point.pdf
• mlp_notation.pdf
• pla_nn_a.pdf
• lda2.pdf

They all seem to be broken, however, there is one

• yaleb_exs.pdf

which can at least be opened in a normal PDF viewer (but it seems to be empty).

[pubpub-zz]

those files are deeply damaged and cannot be opend with acrobat reader (even yaleb_exs ???)
PyPDF2 can not be expecting to read them.

@MartinThoma , this issue should be closed

[pubpub-zz]

+1? (to get below 60😎)

[MartinThoma]

I just tried to open those files. Yes, my PDF Viewer can open some of them (not all), but all of them had just a blank page. The only thing that sometimes changed was the dimension of that blank page.

I'll close this for the moment as we have more important issues to work on that driving robustness up that much.

[MartinThoma]

I've linked it in #1210 as we might want to throw PdfReadError instead of the KeyError.