Windows 7 Pro VM Python 2.7 gives 5 test errors with latest release 0.10.1

[Garyfallidis]

After doing in Windows 7 Pro (64 bit) (Python 2.7, Anaconda 2.0 with Numpy 1.8.1, Cython 0.20.1 and Scipy 0.14.0)

pip install dipy
import dipy
dipy.test()

I get the following errors

======================================================================
ERROR: dipy.segment.tests.test_feature.test_feature_extract
Traceback (most recent call last):
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\nose\case.py", line 197, in runTest
self.test(*self.arg)
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\dipy\segment\tests\test_feature.py", line 231, in test_feature_e
features = extract(feature, streamlines)
File "dipy/segment/featurespeed.pyx", line 371, in dipy.segment.featurespeed.extract (dipy\segment\featurespeed.c:6302
File "dipy/segment/featurespeed.pyx", line 394, in dipy.segment.featurespeed.extract (dipy\segment\featurespeed.c:5973
File "dipy/segment/featurespeed.pyx", line 363, in dipy.segment.featurespeed.infer_shape (dipy\segment\featurespeed.c:
File "dipy/segment/featurespeed.pyx", line 46, in dipy.segment.featurespeed.Feature.c_infer_shape (dipy\segment\featur
7)
TypeError: object of type 'long' has no len()
======================================================================
======================================================================
ERROR: dipy.tracking.tests.test_life.test_voxel2streamline
Traceback (most recent call last):
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\nose\case.py", line 197, in runTest
self.test(*self.arg)
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\dipy\tracking\tests\test_life.py", line 89, in test_voxel2stream
v2f, v2fn = life.voxel2streamline(streamline, False, affine)
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\dipy\tracking\life.py", line 313, in voxel2streamline
return _voxel2streamline(transformed_streamline, unique_idx)
File "dipy/tracking/vox2track.pyx", line 20, in dipy.tracking.vox2track._voxel2streamline (dipy\tracking\vox2track.c:1
ValueError: Buffer dtype mismatch, expected 'npy_intp' but got 'long'
======================================================================
======================================================================
ERROR: dipy.tracking.tests.test_life.test_FiberModel_init
Traceback (most recent call last):
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\nose\case.py", line 197, in runTest
self.test(*self.arg)
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\dipy\tracking\tests\testlife.py", line 111, in test_FiberModel
fiber_matrix, vox_coords = FM.setup(streamline, affine, sphere=sphere)
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\dipy\tracking\life.py", line 377, in setup
affine=affine, unique_idx=vox_coords)
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\dipy\tracking\life.py", line 313, in voxel2streamline
return _voxel2streamline(transformed_streamline, unique_idx)
File "dipy/tracking/vox2track.pyx", line 20, in dipy.tracking.vox2track._voxel2streamline (dipy\tracking\vox2track.c:1
ValueError: Buffer dtype mismatch, expected 'npy_intp' but got 'long'
======================================================================
======================================================================
ERROR: dipy.tracking.tests.test_life.test_FiberFit
Traceback (most recent call last):
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\nose\case.py", line 197, in runTest
self.test(*self.arg)
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\dipy\tracking\tests\test_life.py", line 132, in test_FiberFit
fiber_matrix, vox_coords = FM.setup(streamline, None, evals)
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\dipy\tracking\life.py", line 377, in setup
affine=affine, unique_idx=vox_coords)
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\dipy\tracking\life.py", line 313, in voxel2streamline
return _voxel2streamline(transformed_streamline, unique_idx)
File "dipy/tracking/vox2track.pyx", line 20, in dipy.tracking.vox2track._voxel2streamline (dipy\tracking\vox2track.c:1
*ValueError: Buffer dtype mismatch, expected 'npy_intp' but got 'long'*
======================================================================
======================================================================
FAIL: dipy.viz.tests.test_fvtk_window.test_renderer
Traceback (most recent call last):
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\nose\case.py", line 197, in runTest
self.test(self.arg)
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\nose\util.py", line 619, in newfunc
return func(arg, kw)
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\numpy\testing\decorators.py", line 146, in skipper_func
return f(*args, kwargs)
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\dipy\testing\decorators.py", line 68, in test_with_xvfb
my_test()
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\dipy\viz\tests\test_fvtk_window.py", line 36, in test_renderer
npt.assert_equal(report.colors_found, [True, False])
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\numpy\testing\utils.py", line 255, in assert_equal
assert_equal(actual[k], desired[k], 'item=%r\n%s' % (k, err_msg), verbose)
File "C:\Users\Eleftherios\Anaconda\lib\site-packages\numpy\testing\utils.py", line 317, in assert_equal
raise AssertionError(msg)
AssertionError:
Items are not equal:
item=0
ACTUAL: False
DESIRED: True

[Garyfallidis]

Can someone with a Windows machine try the same? pip install dipy and run the tests? These errors did not appear in the buildbots but appeared in my virtual machine which is a bit unexpected.

[matthew-brett]

Perhaps the error is specific to Anaconda?

[Garyfallidis]

Or Windows 64 bit specific? Can you remind me which of the win bots is the one with 64 bits?
For now I am trying to upgrade cython and recompile.

[matthew-brett]

Could be 64-bits - we don't test Windows 64-bits at the moment.

[Garyfallidis]

We should consider adding one maybe Win 8.1 or 10 machine (64bit) we have nobody working on the core-dev team on windows machines. So, we could be missing some errors there. Maybe Appveyor can help in this regard too.

[matthew-brett]

Yes, sure, I will do that on a spare machine I have in my office on Monday.

[Garyfallidis]

Just a bit of feedback here. Same errors appear with updating Cython.

[Garyfallidis]

Cool, thumbs up for the spare machine!
@matthieudumont and @omarocegueda do you often develop on Windows these days? If yes whenever you see any errors on the nosetests please write an issue or provide a fix if possible. We currently miss eyeballing on 64 bit windows machines. Thx in advance.

[omarocegueda]

Hi!,

@matthieudumont and @omarocegueda do you often develop on Windows these days?

sorry, I have been working on Linux since a few weeks because some automatic windows updates messed up my partition and the hard drive wasn't even recognized, I guess it was a conflict with my dual-boot configuration. Anyway, let me try to repro on my 64bit 8.1.

[omarocegueda]

According to this post:
https://groups.google.com/forum/#!topic/cython-users/UX70kTr-6vk

"There are issues with the mingw runtime conflicting with the MSVC
runtime on Windows 64". This is the kind of errors I get when I attempt to build dipy:

/../x86_64-w64-mingw32/bin/ld.exe: warning: cannot find entry symbol _DllMain@12;

I started seeing these errors when I updated to Anaconda2-2.4-x86-64 in windows 8.1 64 bits. By any chance are you using mingw as well?

I will try using visual c instead (we may need to stop recommending people to use mingw =( )

[matthew-brett]

Omar - I think the buildbots are all using MSVC - certainly for Python 3.5.

[omarocegueda]

Thanks Matthew!, good to know!, I was using mingw because I was having problems with relatively recent versions of MSVC: I was having the typical error message saying that it cannot find vcvarsall.bat even though I had visual studio professional 2012. So at some point I tried mingw, everything worked fine, and I kept using it. Now I see there are more detailed instructions from the cython group.

[omarocegueda]

Alright, I finally succeeded installing anaconda2 and the MS sdk on windows 8 64 bits, and I confirm that many tests are failing. In my case, its 9 regression tests from the registration module.

[matthew-brett]

Attempt at Python 2.7 64-bit builds:

http://nipy.bic.berkeley.edu/builders/dipy-bdist64-27

Attempt at Python 3.5 64-bit builds:

http://nipy.bic.berkeley.edu/builders/dipy-bdist64-35

[matthew-brett]

OK, we now have Python 2.7 and 3.5 builds on Windows 7 64-bit (links above) and a Python 3.5 build on Windows 10:

http://nipy.bic.berkeley.edu/builders/dipy-bdist64-35-win10

[arokem]

Since you have just done this installation, maybe you can help this person on the mailing list?

https://mail.python.org/pipermail/neuroimaging/2015-December/000672.html

[matthew-brett]

That looks to be an Anaconda compilation error - I haven't installed or tested Anaconda. Omar - do you have any insights into that one?

[omarocegueda]

I also had a lot of problems attempting to build extensions on win64. The root of the problem turned out to be mingw. When one looks on the internet, everybody recommends using mingw, but I think that's no longer a good idea because there is no stable version of mingw for 64bit arch. By following these instructions:

https://github.com/cython/cython/wiki/CythonExtensionsOnWindows

everything worked smoothly (except by the long time it took to download the SDK from Microsoft!). I was just hesitant about answering on the mailing list because the error messages this guy mentions are different from the ones I got, I didn't want to provide a misleading advise.

[matthew-brett]

I'm worried that this is actually an Anaconda compiler configuration error.

[arokem]

Do you have instructions somewhere on how to build/install Dipy on Windows
without Anaconda?

On Wed, Dec 9, 2015 at 10:05 PM, Matthew Brett notifications@github.com
wrote:

I'm worried that this is actually an Anaconda compiler configuration
error.

—
Reply to this email directly or view it on GitHub
#803 (comment).

[matthew-brett]

Not specifically, but it's the same for any windows compiled project : https://matthew-brett.github.io/pydagogue/python_msvc.html

[matthew-brett]

I think we should close this one, because I think we have reports +/- proposed fixes for the component failures open now:

#806
#810
#812
#813