Java exception 2016a in Jupyter/IPython over linux

Something regressed between R2015aSP1 and R2016a, when MATLAB is called via Jupyter Notebook using the matlab_kernel python interface.

With R2015aSP1, everything works as expected.

With R2016a, things work except for plotting. When trying to plot anything, we get the error

!!!!!!!!
java.lang.Exception: Stack trace
  at java.lang.Thread.dumpStack(Unknown Source)
  at com.mathworks.hg.util.OutputHelperProcessingException.<init>(OutputHelperProcessingException.java:17)
  at com.mathworks.hg.util.HGRasterOutputHelper.generateOutput(HGRasterOutputHelper.java:170)
!!!!!!!!

after which, the MATLAB instance starts to eat up system RAM and becomes unresponsive. Help?

[Cosmo]

Here's the related issue on GitHub:

https://github.com/arokem/python-matlab-bridge/issues/244

It appears that the developer of pymatbridge is looking into it. Apparently it's something to do with saved figures being too large.

Answer 1

[Adithya Addanki]

Hi Willie,

It may be possible that the graphics drivers are out of date. Please try updating the graphic drivers.

Additionally, please refer to the link below for fixing low-level graphics issues:

http://www.mathworks.com/help/matlab/creating_plots/resolving-low-level-graphics-issues.html (Attention to "Fix Out-of-Memory Issues")

I hope this helps.

Thank you,

Adithya

[Willie Wong]

That's not it.

1. The Xorg graphics drivers are xf86_video_intel version 2.99.917 which is the most recent one.

2. The video is provided by an integrated Intel chip, which has Intel-supported open source drivers in the kernel. I am running the 4.4.5-1 kernel, which was released mid March. So that's also quite very new.

3. I am really doubtful that it is a driver issue: running the MATLAB desktop itself gives no problem what so ever, it is only when MATLAB is called through Jupyter Notebook that causes the problem. And also as I said the 2015aSP1 build gives no problem either, even when called from Jupyter Notebook.