Segmentation fault on saving

[moldyn-nagel]

OS: Debian 10, 4.19
Python: 3.7.3
viscm: 0.9 (master-branch)
matplotlib: 3.1.1
numpy: 1.16.2
colorspacious: 1.1.2

Running viscm in edit mode, I am not able to save it. Either via 'save'-button nor via 'File>Export .py'. It crashes with the statement 'segmentation fault'.
This behavior is independent on how viscm is called via $ python -m viscm, $ python -m viscm edit or $ python -m viscm edit path/to/file.jscm.

[tacaswell]

How did you install python, numpy, and Matplotlib? Getting segmentation faults suggests that something is wrong in the c-extensions.

[moldyn-nagel]

Thank you for your fast reply, python via debian package manager, numpy and matplotlib via python -m pip install --user ... due to missing admin privileges. Btw. python is "using" gcc 8.3.0.

[tacaswell]

Can you reproduce this without viscm (just saving a Matplotlib figure)?

Can you un-install the --user installed Matplotlib and try in a venv instead

If you do

python -m venv /path/to/venv
. /path/to/venv/bin/activate
pip install --upgrade pip
pip install matplotlib
pip install -e path/to/viscm

?

I am concerned there is some ABI miss-match between the debian system packages and Matplotilb wheel.

[moldyn-nagel]

Can you reproduce this without viscm (just saving a Matplotlib figure)?

I do not face any issues there, e.g.

import matplotlib.pyplot as plt
fig = plt.figure()
plt.plot([1, 2, 1])
fig.savefig('test.pdf')

works correctly.

Can you un-install the --user installed Matplotlib and try in a venv instead

I can test it on Monday, python3-venv needs to be reinstalled but the admin already left for weekend.

[moldyn-nagel]

Within the venv viscm seems to work flawless. Thank you for your help!

P.S.: Additional to your commands I needed to install PyQt5 and scipy:

pip install PyQt5
pip install scipy

[tacaswell]

Great, I'm glad that got sorted.

This further re-enforces my view that --user is a miss-feature of pip in the scientific python space (or really anything with complied extensions).