Regression in 1.28.x: get_pixmap() and get_text() hang at 100% CPU on a specific PDF; works in 1.27.2.3

[nyhtn]

Description of the bug

PyMuPDF 1.28.2 hangs indefinitely with approximately 100% CPU usage when processing the first page of a specific publicly available PDF.

Both Page.get_pixmap() and Page.get_text() are affected.

The same PDF works correctly with PyMuPDF 1.27.2.3.

This appears to be a regression introduced between 1.27.2.3 and 1.28.x.

Public PDF:

https://www.city.toyota.aichi.jp/_res/projects/default_project/_page_/001/057/466/01.pdf

PDF size:

2,983,626 bytes

How to reproduce

import urllib.request
import pymupdf

url = "https://www.city.toyota.aichi.jp/_res/projects/default_project/_page_/001/057/466/01.pdf"

with urllib.request.urlopen(url, timeout=30) as response:
    data = response.read()

print("downloaded:", len(data), flush=True)

doc = pymupdf.open(stream=data, filetype="pdf")
page = doc[0]

print("START get_pixmap", flush=True)
page.get_pixmap()
print("END get_pixmap", flush=True)

Run with a timeout, for example:

timeout 10s python repro.py
echo $?

With PyMuPDF 1.28.2, the output is:

downloaded: 2983626
START get_pixmap

The process then consumes approximately 100% CPU and does not return.

The timeout exits with code 124.

get_text() behaves the same way:

page.get_text()

and:

page.get_text("text", flags=0)

Both hang on page 1.

Additional investigation

Page.clean_contents() itself completes successfully:

page.clean_contents()
print("clean complete", flush=True)

page.get_text()

Output:

clean complete

but the subsequent get_text() still hangs.

Therefore, the issue does not appear to be caused only by multiple /Contents streams or text extraction flags.

Rendering is also affected because get_pixmap() hangs, so the issue appears to occur while processing the page contents rather than only during structured text extraction.

Version comparison

PyMuPDF 1.28.2 / MuPDF 1.28.2

PyMuPDF 1.28.2: Python bindings for the MuPDF 1.28.2 library.
Python 3.12 running on linux (64-bit).

Results on page 1:

page.get_pixmap()              -> hangs / high CPU
page.get_text("text", flags=0) -> hangs / high CPU
page.get_text()                -> hangs / high CPU

PyMuPDF 1.27.2.3 / MuPDF 1.27.2

PyMuPDF 1.27.2.3: Python bindings for the MuPDF 1.27.2 library.
Python 3.12 running on linux (64-bit).

The same PDF completes normally.

Example text extraction timings:

PAGE_START 1
PAGE_END 1 elapsed=0.036 chars=959
PAGE_START 2
PAGE_END 2 elapsed=0.014 chars=595
PAGE_START 3
PAGE_END 3 elapsed=0.009 chars=834
PAGE_START 4
PAGE_END 4 elapsed=0.011 chars=740
PAGE_START 5
PAGE_END 5 elapsed=0.007 chars=536
PAGE_START 6
PAGE_END 6 elapsed=0.019 chars=762
PAGE_START 7
PAGE_END 7 elapsed=0.014 chars=840
PAGE_START 8
PAGE_END 8 elapsed=0.019 chars=1449
DONE

Environment

• OS: Linux x86_64
• Python: 3.12
• Installation: official pip wheel
• PyMuPDF: 1.28.2
• MuPDF: 1.28.2
• Wheel: pymupdf-1.28.2-cp310-abi3-manylinux_2_28_x86_64.whl

Expected behavior

The page should render and text extraction should complete as it does with PyMuPDF 1.27.2.3.

Actual behavior

On PyMuPDF 1.28.2, processing page 1 enters a high-CPU state and does not return.

Related issues

The symptoms look similar to previous upstream hangs such as #3125, #3357, #3430, and #4435, but this appears to be a new regression affecting the 1.28.x series and the PDF above.

How to reproduce the bug

I can provide additional diagnostics or test a patched version if needed. Please let me know if there is anything else I can check on my side.

PyMuPDF version

1.28.2

Operating system

Linux

Python version

3.12

[JorjMcKie]

Issue transferred to MuPDF's bug system, https://bugs.ghostscript.com/show_bug.cgi?id=709641