Fitz freezes on some PDFs when calling the fitz.Page.get_text_blocks method.

[AndreyRomanyukov]

Describe the bug (mandatory)

Fitz freezes on some PDFs when calling the fitz.Page.get_text_blocks method.

To Reproduce (mandatory)

Download the pdf that causes a freeze.

original https://aacr.figshare.com/articles/journal_contribution/Supplementary_Data_from_Targeting_Therapeutic_Resistance_and_Multinucleate_Giant_Cells_in_CCNE1-Amplified_HR-Proficient_Ovarian_Cancer/22523824/1/files/39986620.pdf
mirror https://www.dropbox.com/s/s7zjp7a8ys5ibh0/mct-21-0873_supplementary_data_s1_supps1.pdf?dl=0

Run the python code

from io import BytesIO

import fitz


path = 'mct-21-0873_supplementary_data_s1_supps1.pdf'

with open(path, 'rb') as opened:
    stream = BytesIO(opened.read())

with fitz.open(stream=stream, filetype="pdf") as pdf:
    for page in pdf:
        print(page)
        blocks = page.get_text_blocks()
        print('blocks:', len(blocks))

The program will print

page 0 of <memory, doc# 1>
blocks: 2
page 1 of <memory, doc# 1>
blocks: 2
page 2 of <memory, doc# 1>

and then freeze.

Additional context (optional)

Reproduces on PyMuPDF==1.22.3 and PyMuPDF==1.22.5. Reproduces on macOS 12.6.5 and Ubuntu 20.04.2

[JorjMcKie]

The problem occurs when extracting text in general. It more precisely is loop.
It is also not a problem inside PyMuPDF but in the base library MuPDF.
mutool draw -o test.txt file.pdf also loops.
I will forward the problem to MuPDF's issue system.

[AndreyRomanyukov]

I will forward the problem to MuPDF's issue system.

Thanks!

[JorjMcKie]

Sorry for the delay, here is the MuPDF bug report: https://bugs.ghostscript.com/show_bug.cgi?id=707074.

[JorjMcKie]

The problem has been fixed in MuPDF.
The file is damaged in the sense that it contains circular references within its objects - that hadn't been detected before.
Now the circular references are detected and your script with raise an exception - as it should.

[julian-smith-artifex-com]

Fixed in 1.23.5.