Discussion:
[Matplotlib-users] Fw: Matplotlib 1.5.1 - Function pcolormesh throws Segmentation Fault / faulthandler traceback info / Error at Agg
Claude Falbriard
2016-02-26 21:20:57 UTC
Permalink
Dear colleagues,

Complementing this topic:

. The segmentation fault error relates to the use of the "Agg" graphical
back-end which crashes with larger plots when using the command pyplot
savefig.
Case I switch to the .use("Cairo") package, the processing completes OK
although the image is not matching the color scheme.
This machine, a z13, does not have an interactive X display, its a cpu
in a cloud, a back-end machine with web access.

My question 1:
Is it possible to recompile the failing "Agg" framework from its source in
"c" and make it compatible with the hardware. Any hints how to run the
make
build and install and add it to the Matplotlib pyplot services?

My question 2:
Is there an alternate back-end solution available for .png or .tif,
preferably in a Python space, that is accepted by the matplotlib pyplot
tool as
a valid choice and does not require a display configuration? Worst case,
is there a bypass solution available?

Thanks advance for your guidance.

Regards,
Claude

History:
----------

Thanks for the quick feedback.

I've installed the "faulthandler" package, added it to the import and
enabled it with faulthandler.enable().
Below the console trace of the test using the pcolormesh function and the
traceback information:

python test3.py

Fatal Python error: Segmentation fault

Current thread 0x000003ffa45f7710 (most recent call first):
File "/usr/lib64/python2.7/site-packages/matplotlib/collections.py",
line 1877 in draw
File "/usr/lib64/python2.7/site-packages/matplotlib/artist.py", line 61
in draw_wrapper
File "/usr/lib64/python2.7/site-packages/matplotlib/axes/_base.py", line
2324 in draw
File "/usr/lib64/python2.7/site-packages/matplotlib/artist.py", line 61
in draw_wrapper
File "/usr/lib64/python2.7/site-packages/matplotlib/figure.py", line
1159 in draw
File "/usr/lib64/python2.7/site-packages/matplotlib/artist.py", line 61
in draw_wrapper
File
"/usr/lib64/python2.7/site-packages/matplotlib/backends/backend_agg.py",
line 474 in draw
File
"/usr/lib64/python2.7/site-packages/matplotlib/backends/backend_agg.py",
line 527 in print_png
File "/usr/lib64/python2.7/site-packages/matplotlib/backend_bases.py",
line 2232 in print_figure
File "/usr/lib64/python2.7/site-packages/matplotlib/figure.py", line
1565 in savefig
File "/usr/lib64/python2.7/site-packages/matplotlib/pyplot.py", line 688
in savefig
File "test3.py", line 47 in <module>
Segmentation fault

Hope this information helps to circle the issue.


Regards,

Claude Falbriard
Certified IT Specialist L2 - Middleware


Phone: 55-13-99662-5703 | Mobile: 55-13-98117-3316
E-mail: ***@br.ibm.com




From: Benjamin Root <***@gmail.com>
To: Claude Falbriard/Brazil/***@IBMBR
Cc: Matplotlib Users <matplotlib-***@lists.sourceforge.net>
Date: 24/02/2016 17:35
Subject: Re: [Matplotlib-users] Matplotlib 1.5.1 - Function
pcolormesh throws Segmentation Fault



Sorry, forgot to post the link: https://pypi.python.org/pypi/faulthandler/

On Wed, Feb 24, 2016 at 3:33 PM, Benjamin Root <***@gmail.com>
wrote:
Could you try using faulthandler and post the traceback please? That'll
help us isolate the problem better.

Ben Root

On Wed, Feb 24, 2016 at 3:04 PM, Claude Falbriard <***@br.ibm.com>
wrote:
Dear colleagues,

I've done a build from source of latest Matplotlib package and deployed
it at our IBM z13 machine (s390x). It uses the current release 1.5.1.
During the unit tests I found an issue with a test case from NOAA which
uses a pcolormesh draw function with basemap.

Example 2: Plot data from an NWW3 GRiB2 file - [ here:
http://polar.ncep.noaa.gov/waves/examples/usingpython.shtml]

The following line is causing a Segmentation fault error even when adding
an 8GB swap memory to the process:

cs = m.pcolormesh(x,y,data,shading='flat',cmap=plt.cm.jet)

I also tryed to execute other, similar samples that use pcolormesh, but
receiving the same error. Is this a known issue or might it be be related
to the memory environment ? Any hints how to debug this error?

Regards,

Claude Falbriard
Certified IT Specialist L2 - Middleware


Phone:55-13-99662-5703 | Mobile:55-13-98117-3316
E-mail: ***@br.ibm.com

Loading...