failing test "test_imresize4 (test_pilutil.TestPILUtil)" of scipy1.4.1 with pillow2.7.0 (on windows)

[stonebig]

with latest @cgohlke wheelaries for pillow2.7.0 and scipy0.14.1, I have still the issue on windows 64bit python3.4.2 (and also 32bit).... when trying to check if my installation is ok

import scipy;scipy.test()

======================================================================
FAIL: test_imresize4 (test_pilutil.TestPILUtil)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "D:\result_tests\WinPython-64bit-3.4.2.4_build5FlavorRfull\python-3.4.2.amd64\lib\site-packages\numpy\testing\decorators.py", line 146, in skipper_func
    return f(*args, **kwargs)
  File "D:\result_tests\WinPython-64bit-3.4.2.4_build5FlavorRfull\python-3.4.2.amd64\lib\site-packages\scipy\misc\tests\test_pilutil.py", line 57, in test_imresize4
    assert_equal(im2, res)
  File "D:\result_tests\WinPython-64bit-3.4.2.4_build5FlavorRfull\python-3.4.2.amd64\lib\site-packages\numpy\testing\utils.py", line 275, in assert_equal
    return assert_array_equal(actual, desired, err_msg, verbose)
  File "D:\result_tests\WinPython-64bit-3.4.2.4_build5FlavorRfull\python-3.4.2.amd64\lib\site-packages\numpy\testing\utils.py", line 739, in assert_array_equal
    verbose=verbose, header='Arrays are not equal')
  File "D:\result_tests\WinPython-64bit-3.4.2.4_build5FlavorRfull\python-3.4.2.amd64\lib\site-packages\numpy\testing\utils.py", line 665, in assert_array_compare
    raise AssertionError(msg)
AssertionError: 
Arrays are not equal

(mismatch 75.0%)
 x: array([[ 1.  ,  1.25,  1.75,  2.  ],
       [ 1.5 ,  1.75,  2.25,  2.5 ],
       [ 2.5 ,  2.75,  3.25,  3.5 ],
       [ 3.  ,  3.25,  3.75,  4.  ]], dtype=float32)
 y: array([[ 1. ,  1. ,  1.5,  2. ],
       [ 1. ,  1. ,  1.5,  2. ],
       [ 2. ,  2. ,  2.5,  3. ],
       [ 3. ,  3. ,  3.5,  4. ]], dtype=float32)

[homm]

This is obviously very bad scipy developers idea to test third-party functionality in it's own unit tests. This is wrong result of bilinear interpolation produced by previous Pillow and PIL version. This is definitely scipy issue, not Pillow.

[wiredfool]

I ran across this @ release time and meant to get a PR to scipy.

This test has exposed regressions before, so we actually have a copy of it in our tree as well. The comments on the test say that they're testing for the same return value for various different ways of specifying the resize size target. They're just testing against a hardwired (and wrong) array rather than just making sure that Pillow is doing the same thing for different inputs at their end.

[wiredfool]

Ok PR sent.

[stonebig]

thanks !

[wiredfool]

It's been merged in Scipy.