Could not Initialize GLX after Conda Update --All

Problem Description

I’m using Spyder on a remote server by SSH and X11 forwarding. After I use conda update --all , spyder stopped launching and has the error: Could not initialize GLX. I have tried adding these lines in start.py

import ctypes
ctypes.CDLL("libGL.so.1", mode=ctypes.RTLD_GLOBAL)
from PyQt5 import QtWebEngineWidgets

and also

export __GLX_VENDOR_LIBRARY_NAME=nvidia

but none of them works.

I also tried

conda install qt=5.6.2

but then I encounter the segmentation fault.

What steps reproduce the problem?

1. Create a new environment
2. Install keras, keras gpu, tensorflow, tensorflow gpu, cudatoolkit and dependencies
3. Conda update --all

Versions

• Spyder version: 3.2.8
• Python version: 3.6
• Qt version: 5.9.5
• PyQt version: 5.9.2
• Operating System name/version: CentOS 7

[arnonerba]

This does not appear to be fixed in Spyder 3.3.0. Could this be a Qt issue as noted in #7447?

conda install qt=5.6* pyqt=5.6* still works as a temporary fix with Spyder 3.3.0.

[kentaroy47]

Spyder fails with the same error with Spyder 3.3.1. However, by upgrading conda to the newest version, Spyder 3.3.3 worked ok with ssh -X connections

Also found that using TightVNC causes this error and using vnc4server evades the problem.