I just installed 1.4.2-3.1 on my linux-x86_64 machine, and there the imports work well:
import matplotlib._path and import matplotlib.image as mimg etc. (everything that was in the exception trace). Therefore, I would guess that it works with 1.4.2. On kfreebsd, I also found the .so file (/usr/lib/python3/dist-packages/matplotlib/_path.cpython-34m-x86_64-kfreebsd-gnu.so); however I couldn't test it. Does the imports above work well with them? Best Olr On 03.01.2016 14:16, Steven Chamberlain wrote: > Hi, > > Ole Streicher wrote: >> To me, this looks more that matplotlib-1.4.2 is buggy on kfreebsd. The >> script in question is unchanged since a year, and worked well with >> matplotlib 1.4.2 on linux-amd64 and others. > > I don't know for sure, maybe I need to look at this more closely first. > > python-astropy/1.0.3-1 built successfully with > python-matplotlib/1.4.2-3.1 : > https://buildd.debian.org/status/fetch.php?pkg=python-astropy&arch=kfreebsd-amd64&ver=1.0.3-3&stamp=1437054460 > > The regression started with python-astropy/1.0.5-1, when we still had > the same python-matplotlib/1.4.2-3.1 on kfreebsd : > > | import numpy as np > | from numpy import ma > | > from matplotlib._path import (affine_transform, > | > count_bboxes_overlapping_bbox, > | update_path_extents) > | E ImportError: No module named 'matplotlib._path' > https://buildd.debian.org/status/fetch.php?pkg=python-astropy&arch=kfreebsd-amd64&ver=1.0.5-2&stamp=1444321350 > > And now with python-astropy/1.0.5-1 on kfreebsd, we gave back > python-matplotlib/1.4.2-3.1 for another built attempt and it built. > That's how I came to this conclusion, but maybe there's another reason. > > Regards, >
