Shading tests failing on master

[tacaswell]

======================================================================
FAIL: Array comparison test for the default "hsv" blend mode. Ensure the
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/virtualenv/python2.6.9/lib/python2.6/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/travis/virtualenv/python2.6.9/lib/python2.6/site-packages/matplotlib-1.5.x-py2.6-linux-x86_64.egg/matplotlib/tests/test_colors.py", line 304, in test_light_source_shading_default
    assert_array_almost_equal(rgb, expect, decimal=2)
  File "/home/travis/virtualenv/python2.6.9/lib/python2.6/site-packages/numpy/testing/utils.py", line 842, in assert_array_almost_equal
    precision=decimal)
  File "/home/travis/virtualenv/python2.6.9/lib/python2.6/site-packages/numpy/testing/utils.py", line 665, in assert_array_compare
    raise AssertionError(msg)
AssertionError: 
Arrays are not almost equal to 2 decimals

(mismatch 16.40625%)
 x: array([[[  8.72e-01,   8.72e-01,   8.72e-01,   1.00e+00],
        [  8.31e-01,   7.63e-01,   7.20e-01,   1.00e+00],
        [  8.75e-01,   7.90e-01,   7.38e-01,   1.00e+00],...
 y: array([[[ 0.87,  0.87,  0.87,  1.  ],
        [ 0.85,  0.79,  0.75,  1.  ],
        [ 0.9 ,  0.83,  0.78,  1.  ],...

======================================================================
FAIL: Array comparison test for a surface with a masked portion. Ensures that
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/virtualenv/python2.6.9/lib/python2.6/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/travis/virtualenv/python2.6.9/lib/python2.6/site-packages/matplotlib-1.5.x-py2.6-linux-x86_64.egg/matplotlib/tests/test_colors.py", line 355, in test_light_source_masked_shading
    assert_array_almost_equal(rgb, expect, decimal=2)
  File "/home/travis/virtualenv/python2.6.9/lib/python2.6/site-packages/numpy/testing/utils.py", line 842, in assert_array_almost_equal
    precision=decimal)
  File "/home/travis/virtualenv/python2.6.9/lib/python2.6/site-packages/numpy/testing/utils.py", line 665, in assert_array_compare
    raise AssertionError(msg)
AssertionError: 
Arrays are not almost equal to 2 decimals

(mismatch 29.296875%)
 x: array([[[  0.00e+00,   0.00e+00,   0.00e+00,   1.00e+00],
        [  9.50e-01,   9.27e-01,   9.13e-01,   1.00e+00],
        [  5.33e-01,   3.37e-01,   2.15e-01,   1.00e+00],...
 y: array([[[ 1.  ,  1.  ,  1.  ,  1.  ],
        [ 0.95,  0.93,  0.91,  1.  ],
        [ 0.96,  0.93,  0.91,  1.  ],...

----------------------------------------------------------------------

I get the same failures locally.

cc @jenshnielsen @joferkington

[tacaswell]

I think the failing tests were added in #3291

[tacaswell]

This is related to numpy 1.8 v 1.9

(conda27)tcaswell@eowyn:matplotlib$ python tests.py matplotlib.tests.test_colors
................
----------------------------------------------------------------------
Ran 16 tests in 3.639s

OK
(conda27)tcaswell@eowyn:matplotlib$ conda upgrade numpy
Error: Could not locate 'conda-upgrade'

Did you mean one of these?
    update
(conda27)tcaswell@eowyn:matplotlib$ conda update numpy 
Fetching package metadata: ..
Solving package specifications: .
Package plan for installation in environment /home/tcaswell/.virtualenvs/conda27:

The following packages will be UPDATED:

    numpy: 1.8.2-py27_0 --> 1.9.0-py27_0

Proceed ([y]/n)? y

Unlinking packages ...
[      COMPLETE      ] |##########################################################################################################################| 100%
Linking packages ...
[      COMPLETE      ] |##########################################################################################################################| 100%
(conda27)tcaswell@eowyn:matplotlib$ python tests.py matplotlib.tests.test_colors
............FF..
======================================================================
FAIL: Array comparison test for the default "hsv" blend mode. Ensure the
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/tcaswell/.virtualenvs/conda27/lib/python2.7/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/tcaswell/other_source/matplotlib/lib/matplotlib/tests/test_colors.py", line 304, in test_light_source_shading_default
    assert_array_almost_equal(rgb, expect, decimal=2)
  File "/home/tcaswell/.virtualenvs/conda27/lib/python2.7/site-packages/numpy/testing/utils.py", line 842, in assert_array_almost_equal
    precision=decimal)
  File "/home/tcaswell/.virtualenvs/conda27/lib/python2.7/site-packages/numpy/testing/utils.py", line 665, in assert_array_compare
    raise AssertionError(msg)
AssertionError: 
Arrays are not almost equal to 2 decimals

(mismatch 16.40625%)
 x: array([[[  8.72e-01,   8.72e-01,   8.72e-01,   1.00e+00],
        [  8.31e-01,   7.63e-01,   7.20e-01,   1.00e+00],
        [  8.75e-01,   7.90e-01,   7.38e-01,   1.00e+00],...
 y: array([[[ 0.87,  0.87,  0.87,  1.  ],
        [ 0.85,  0.79,  0.75,  1.  ],
        [ 0.9 ,  0.83,  0.78,  1.  ],...

======================================================================
FAIL: Array comparison test for a surface with a masked portion. Ensures that
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/tcaswell/.virtualenvs/conda27/lib/python2.7/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/tcaswell/other_source/matplotlib/lib/matplotlib/tests/test_colors.py", line 355, in test_light_source_masked_shading
    assert_array_almost_equal(rgb, expect, decimal=2)
  File "/home/tcaswell/.virtualenvs/conda27/lib/python2.7/site-packages/numpy/testing/utils.py", line 842, in assert_array_almost_equal
    precision=decimal)
  File "/home/tcaswell/.virtualenvs/conda27/lib/python2.7/site-packages/numpy/testing/utils.py", line 665, in assert_array_compare
    raise AssertionError(msg)
AssertionError: 
Arrays are not almost equal to 2 decimals

(mismatch 29.296875%)
 x: array([[[  0.00e+00,   0.00e+00,   0.00e+00,   1.00e+00],
        [  9.50e-01,   9.27e-01,   9.13e-01,   1.00e+00],
        [  5.33e-01,   3.37e-01,   2.15e-01,   1.00e+00],...
 y: array([[[ 1.  ,  1.  ,  1.  ,  1.  ],
        [ 0.95,  0.93,  0.91,  1.  ],
        [ 0.96,  0.93,  0.91,  1.  ],...

----------------------------------------------------------------------
Ran 16 tests in 3.076s

FAILED (failures=2)

cc @mwiebe

[joferkington]

I think there was a change to the default behavior to numpy.gradient in 1.9. I'm guessing, but these are probably related to that change.

[jenshnielsen]

This is a duplicate of #3598 and yes I think this is due to the gradient change in numpy. I

[tacaswell]

Sorry for the noise, despite commenting on that other issue I had no memory of it..

[jenshnielsen]

No problem, I should have had a fix for that thing long ago, but have been to busy with other stuff and pushed this back since it is not relevant for the 1.4.1 release.