AssertionError: assert 0 < size <= self._size in tornado/iostream.py when displaying a large image before other images

[kerrickstaley]

Summary

When I try to display a large-ish image (800x800 JPEG) using the Pillow Python library in Jupyter Notebook before displaying any smaller images in the notebook, the notebook hangs. In the terminal, I see this error message:

[E 11:01:18.952 NotebookApp] Uncaught exception, closing connection.
    Traceback (most recent call last):
      File "/Users/kerrick/Library/Python/3.11/lib/python/site-packages/tornado/iostream.py", line 702, in _handle_events
        self._handle_write()
      File "/Users/kerrick/Library/Python/3.11/lib/python/site-packages/tornado/iostream.py", line 976, in _handle_write
        self._write_buffer.advance(num_bytes)
      File "/Users/kerrick/Library/Python/3.11/lib/python/site-packages/tornado/iostream.py", line 182, in advance
        assert 0 < size <= self._size
               ^^^^^^^^^^^^^^^^^^^^^^
    AssertionError
Exception in callback None()
handle: <Handle cancelled>
Traceback (most recent call last):
  File "/opt/homebrew/Cellar/python@3.11/3.11.4/Frameworks/Python.framework/Versions/3.11/lib/python3.11/asyncio/events.py", line 80, in _run
    self._context.run(self._callback, *self._args)
  File "/Users/kerrick/Library/Python/3.11/lib/python/site-packages/tornado/platform/asyncio.py", line 206, in _handle_events
    handler_func(fileobj, events)
  File "/Users/kerrick/Library/Python/3.11/lib/python/site-packages/tornado/iostream.py", line 702, in _handle_events
    self._handle_write()
  File "/Users/kerrick/Library/Python/3.11/lib/python/site-packages/tornado/iostream.py", line 976, in _handle_write
    self._write_buffer.advance(num_bytes)
  File "/Users/kerrick/Library/Python/3.11/lib/python/site-packages/tornado/iostream.py", line 182, in advance
    assert 0 < size <= self._size
           ^^^^^^^^^^^^^^^^^^^^^^
AssertionError

This doesn't always happen consistently. Sometimes the image displays successfully the first time but then I get the crash when I re-run the cell. Other times it displays halfway and cuts off in the middle.

A 600x600 rescale of the same image does not crash the notebook. Also, interestingly, if I display the 600x600 image first and then display the 800x800 image in a later cell, it does not crash.

After this happens the notebook is unresponsive and I can't seem to restart the notebook kernel from the UI; the only recourse is to Ctrl+C the python3 -m jupyter notebook process from the terminal.

To reproduce

1. Download crash_800x800.jpg and no_crash_600x600.jpg from this Google Drive folder. (Try clicking "Download all" if the normal "Download" link doesn't work; Google seems to have a bug).
2. Install the Pillow Python library.
3. Run this code in a notebook. Maybe re-run the cell a couple times:

from PIL import Image
Image.open('crash_800x800.jpg')

Expected behavior

The image displays.

Actual behavior

I get a crash with the above-mentioned error message.

Version information

OS: macOS 13.3.1 (a) (22E772610a) on MacBook Air M2
Jupyter Version: 6.5.4-2c95aa495
Jupyter Python version: Python 3.11.4 (main, Jun 7 2023, 00:34:59) [Clang 14.0.3 (clang-1403.0.22.14.1)]
Jupyter Kernel:

Python 3.11.4 (main, Jun  7 2023, 00:34:59) [Clang 14.0.3 (clang-1403.0.22.14.1)]
IPython 8.12.0

Browser: Chrome 112.0.5615.137 (Official Build) (arm64)

Notes:

If you instead use no_crash_600x600.jpg, the crash will not trigger.

Also, if you run these two cells in order:

from PIL import Image
Image.open('no_crash_600x600.jpg')

Image.open('crash_800x800.jpg')

then you can re-run the second cell repeatedly and it will not crash.

[RRosio]

Hi @kerrickstaley thank you for submitting this issue. This is a known bug that has been reported in #6721. Closing this issue as a duplicate of that one.