Repository navigation
All attempts to plot fail with "OverflowError: Python int too large to convert to C long" #7633
Description
To help us understand and resolve your issue please check that you have provided
the information below.
- Matplotlib version, Python version and Platform (Windows, OSX, Linux ...)
Platform: Debian Testing, using the python 2.7 (2.7.11-2 when I first started seeing this, just upgraded to 2.7.13rc1 but the problem still exists). I have this problem with the 1.5.3 on debian testing, but I also tried a virtualenv with pip install --pre matplotlib, which pulled in 2.0.0rc1, and I see the same problem.
Debian stable, with matplotlib 1.4.2, works fine, as did Debian testing until recently.
- How did you install Matplotlib and Python (pip, anaconda, from source ...)
See above: Debian package and virtualenv.
- If possible please supply a Short, Self Contained, Correct, Example
that demonstrates the issue i.e a small piece of code which reproduces the issue
and can be run with out any other (or as few as possible) external dependencies.
Minimal test program:
import pylab
distances = [0, 0.0011970407617050934, 0.0050796062254631295, 0.007303941153223762, 0.007303941153223762, 0.008232870398475445, 0.008232870398475445, 0.012806461740481582, 0.016651523629144134, 0.021047330616832543]
elevations = [6404.2, 6410.76, 6400.92, 6410.76, 6423.88, 6423.88, 6453.41, 6410.76, 6427.17, 6420.6]
pylab.plot(distances, elevations)
pylab.show()
(Please let me know if you want a non-pylab example. I see it with straight matplotlib too.)
- If this is an image generation bug attach a screenshot demonstrating the issue.
- If this is a regression (Used to work in an earlier version of Matplotlib), please
note where it used to work.
It worked as recently as a month ago on Debian testing, but sorry, I don't know what matplotlib version that was since I didn't know it was going to stop working.
Activity
efiring commentedon Dec 18, 2016
Would you provide a full traceback, please? What hardware are you running on? Is the system 32-bit or 64-bit?
akkana commentedon Dec 18, 2016
Sorry, of course I should have provided that.
Exception in Tkinter callback
Traceback (most recent call last):
File "/usr/lib/python2.7/lib-tk/Tkinter.py", line 1545, in call
return self.func(*args)
File "/usr/lib/python2.7/dist-packages/matplotlib/backends/backend_tkagg.py", line 283, in resize
self.show()
File "/usr/lib/python2.7/dist-packages/matplotlib/backends/backend_tkagg.py", line 355, in draw
tkagg.blit(self._tkphoto, self.renderer._renderer, colormode=2)
File "/usr/lib/python2.7/dist-packages/matplotlib/backends/tkagg.py", line 26, in blit
_tkagg.tkinit(tk.interpaddr(), 1)
OverflowError: Python int too large to convert to C long
akkana commentedon Dec 18, 2016
The system is 32-bit. The hardware is a Lenovo X201.
akkana commentedon Dec 18, 2016
I just tried setting up a venv where I downgraded all the way to 1.4.2 (since I know it works there on debian stable) but that doesn't tell me much; it doesn't hit the overflow error, but it exits without showing a window. The same thing happens if I try 1.5.0 or 1.5.1. Maybe there's something in the virtualenv that matplotlib needs and isn't getting? (I'm pretty new to virtualenvs.) 1.5.2 is the first version where I see the OverflowError. It might also be worth mentioning that in 1.5.2, 1.5.3 and the --pre version (sorry, didn't record it, I think it was 2.0-something?) where it prints the error, it shows a window but the window is blank. In these earlier versions inside the venv, I don't get a window, it just exits. In Debian Stable 1.4.2, everything works as expected, a window comes up with a plot in it.
QuLogic commentedon Dec 18, 2016
The TkAgg backend is using an integer to hold a pointer; this is unfortunate and bound to cause trouble, except it will be rather random as it depends on whether the pointer is too large or not. On 32-bit systems, this will occur quite a bit more often since integers are smaller.
Here is a related issue in Pillow, except that a long time ago, the type was changed from an int to a long.
Matplotlib is still using an int though, so that needs to be fixed.
Use correct type for Tk addresses.
cgohlke commentedon Dec 18, 2016
FWIW, I think it's issues like these (extensions storing pointers in signed integers etc) why the official 32-bit Python distributions for Windows do not allow to use the address space above 2 GB. See http://bugs.python.org/issue1449496.
sandrotosi commentedon Dec 18, 2016
@tacaswell do you plan to include this in 2.0.0rc2?
tacaswell commentedon Dec 18, 2016
@sandrotosi Yes, I have sunk most of my morning into understanding this will enough to merge @QuLogic 's patch 😉 to merge it and then tag rc2.
Does debian do any patching of python's tk modules?
sandrotosi commentedon Dec 18, 2016
nice thanks!
i dont see any patch to the tk modules in the python2.7 interpreter packaging repo