Crash with Variable Explorer when opening inherited class from `set` in Python 3.13

[tristanmuzzy]

Description

What steps will reproduce the problem?

Using Python 3.13.2 install, not internal Spyder, not sure if relvant

• Spyder version: 6.0.4 (standalone)
• Python version: 3.11.11 64-bit
• Qt version: 5.15.8
• PyQt5 version: 5.15.9
• Operating System: Windows-10-10.0.19045-SP0

run this code

class my_set(set):
    def __init__(self, *args):
        super().__init__(*args) ## set has no kwargs, *args initalizes inherited set
        
        
ms = my_set()

Click on "ms" in Variable Explorer, then Spyder crashes

Versions

• Spyder version: 6.0.4 (standalone)
• Python version: 3.11.11 64-bit
• Qt version: 5.15.8
• PyQt5 version: 5.15.9
• Operating System: Windows-10-10.0.19045-SP0

Dependencies

# Mandatory:
aiohttp >=3.9.3                  :  3.11.12 (OK)
asyncssh >=2.14.0,<3.0.0         :  2.19.0 (OK)
atomicwrites >=1.2.0             :  1.4.1 (OK)
chardet >=2.0.0                  :  5.2.0 (OK)
cloudpickle >=0.5.0              :  3.1.1 (OK)
cookiecutter >=1.6.0             :  2.6.0 (OK)
diff_match_patch >=20181111      :  20241021 (OK)
github >=2.3.0                   :  2.5.0 (OK)
importlib_metadata >=4.6.0       :  8.6.1 (OK)
intervaltree >=3.0.2             :  3.1.0 (OK)
IPython >=8.13.0,<9.0.0,!=8.17.1 :  8.32.0 (OK)
jedi >=0.17.2,<0.20.0            :  0.19.2 (OK)
jellyfish >=0.7                  :  1.1.3 (OK)
jsonschema >=3.2.0               :  4.23.0 (OK)
keyring >=17.0.0                 :  25.6.0 (OK)
nbconvert >=4.0                  :  7.16.6 (OK)
numpydoc >=0.6.0                 :  1.8.0 (OK)
parso >=0.7.0,<0.9.0             :  0.8.4 (OK)
pexpect >=4.4.0                  :  4.9.0 (OK)
pickleshare >=0.4                :  0.7.5 (OK)
psutil >=5.3                     :  6.1.1 (OK)
pygments >=2.0                   :  2.19.1 (OK)
pylint >=3.1,<4                  :  3.3.4 (OK)
pylint_venv >=3.0.2              :  3.0.4 (OK)
pyls_spyder >=0.4.0              :  0.4.0 (OK)
pylsp >=1.12.0,<1.13.0           :  1.12.1 (OK)
pylsp_black >=2.0.0,<3.0.0       :  2.0.0 (OK)
pyuca >=1.2                      :  1.2 (OK)
qdarkstyle >=3.2.0,<3.3.0        :  3.2.3 (OK)
qstylizer >=0.2.2                :  0.2.4 (OK)
qtawesome >=1.3.1,<1.4.0         :  1.3.1 (OK)
qtconsole >=5.6.1,<5.7.0         :  5.6.1 (OK)
qtpy >=2.4.0                     :  2.4.2 (OK)
rtree >=0.9.7                    :  1.3.0 (OK)
setuptools >=49.6.0              :  75.8.0 (OK)
sphinx >=0.6.6                   :  8.1.3 (OK)
spyder_kernels >=3.0.3,<3.1.0    :  3.0.3 (OK)
superqt >=0.6.2,<1.0.0           :  0.7.0 (OK)
textdistance >=4.2.0             :  4.6.2 (OK)
three_merge >=0.1.1              :  0.1.1 (OK)
watchdog >=0.10.3                :  6.0.0 (OK)
yarl >=1.9.4                     :  1.18.3 (OK)
zmq >=24.0.0                     :  26.2.1 (OK)

# Optional:
cython >=0.21                    :  3.0.11 (OK)
matplotlib >=3.0.0               :  3.10.0 (OK)
numpy >=1.7                      :  2.2.2 (OK)
pandas >=1.1.1                   :  2.2.3 (OK)
scipy >=0.17.0                   :  1.15.1 (OK)
sympy >=0.7.3                    :  1.13.3 (OK)

Environment

Environment

Package               Version     Editable project location
--------------------- ----------- ----------------------------------------------------------------
args                  0.1.0
asteval               1.0.6
asttokens             3.0.0
cffi                  1.17.1
clint                 0.5.1
cloudpickle           3.1.1
clr                   1.0.3
clr_loader            0.2.7.post0
cmmr_src              0.0.0       C:\Users\trist\OneDrive\Desktop\Research\CMMR_private\python_pip
colorama              0.4.6
comm                  0.2.2
contourpy             1.3.1
coverage              7.7.1
cycler                0.12.1
debugpy               1.8.13
decorator             5.2.1
dill                  0.4.0
example_package_muzzy 0.0.1       C:\Users\trist\OneDrive\Desktop\Research\package_tutorial
executing             2.2.0
fonttools             4.56.0
ipykernel             6.29.5
ipython               8.34.0
jedi                  0.19.2
jupyter_client        8.6.3
jupyter_core          5.7.2
kiwisolver            1.4.8
llvmlite              0.44.0
lmfit                 1.3.3
mamba                 0.11.3
matplotlib            3.10.1
matplotlib-inline     0.1.7
mpltern               1.0.4
nest-asyncio          1.6.0
numba                 0.61.2
numpy                 2.2.3
packaging             24.2
parso                 0.8.4
pillow                11.1.0
pip                   25.0.1
platformdirs          4.3.6
prompt_toolkit        3.0.50
psutil                7.0.0
pure_eval             0.2.3
pycparser             2.22
Pygments              2.19.1
pyparsing             3.2.1
PyQt6                 6.8.1
PyQt6-Qt6             6.8.2
PyQt6_sip             13.10.0
PySide6_Essentials    6.8.2.1
python-dateutil       2.9.0.post0
pythonnet             3.0.5
pywin32               308
pyzmq                 26.2.1
scipy                 1.15.2
shiboken6             6.8.2.1
six                   1.17.0
spyder-kernels        3.0.3
stack-data            0.6.3
tornado               6.4.2
tqdm                  4.67.1
traitlets             5.14.3
uncertainties         3.2.3
wcwidth               0.2.13
WinTmp                1.2.0
WMI                   1.5.1

[jitseniesen]

I reproduced this on Linux with Spyder master branch running on Python 3.12.7 and the console running on either Python 3.13.3 or 3.13.0. Opening the variable ms leads to a segmentation fault. Opening an empty set, set(), works fine. Running the consle on Python 3.12.10 also works fine. I did not investigate further. No idea about the cause.

[jitseniesen]

The issue seems to be inherent in cloudpickle. If you pickle the ms object in Python 3.13 and unpickle it in Python 3.12, then a segmentation fault ensues (details below). Actually, cloudpickle is not supposed to be used like this. Its documentation at https://github.com/cloudpipe/cloudpickle says:

Cloudpickle can only be used to send objects between the exact same version of Python.

But in Spyder we use cloudpickle to send objects from the console environment to Spyder's runtime environment. The only solution I can think of is to replace cloudpickle by another library for serializing Python objects.

To reproduce the problem with cloudpickle, run the following script in Python 3.13:

import cloudpickle

class my_set(set):
    def __init__(self, *args):
        super().__init__(*args) ## set has no kwargs, *args initalizes inherited set

ms = my_set()

with open('myset-pickle', 'wb') as outfile:
    cloudpickle.dump(ms, outfile)

This serializes ms and write it to the file "myset-pickle". Then, run the following commands in Python 3.12 in the same directory to (try to) read the object from this file:

import cloudpickle
ms = cloudpickle.load(open('myset-pickle', 'rb'))

This leads to a segmentation fault (at least on my Linux computer, and very likely on the Windows computer of the original poster). I used Python versions 3.13.5 and 3.12.8 and cloudpickle version 3.1.1 to test.

[ccordoba12]

Its documentation at https://github.com/cloudpipe/cloudpickle says:

Cloudpickle can only be used to send objects between the exact same version of Python.

Yep, I was aware of that but after reviewing Cloudpickle's code I saw that serialized objects are mostly compatible between Python versions.

The only solution I can think of is to replace cloudpickle by another library for serializing Python objects.

You could explore how dill works (that's the only other widely used Python serialization library I'm aware of). If it solves this problem, then we could switch to it for 6.2.