I will look at it later today, using Steve's amazon machine.

Library issue is fixed in this commit: (tested by Steve on Windows 2008)

http://viewvc.slicer.org/viewvc.cgi/Slicer4?view=revision&revision=19354

The issue of vtk.numpy_support is still there, but it's outside my domain. From Steve's explanation, hopefully it will not cause issues on Windows release build.

The issue with numpy on windows debug builds still exists so I'm switching this over to Jc.

From Jc Email - March 27:

It could be a Debug vs Release issue specific to numpy.

What happen if you do the following:

./Slicer --disable-cli-modules --disable-loadable-modules --disable-scripted-loadable-modules --no-main-window --show-python-interactor

Then:
import numpy.core.numeric
numpy.core.numeric.multiarray.file

The last command should something similar to:
'C:\D\N\Slicer-build-0\python-build\lib\site-packages\numpy\core\numeric\multiarray.pyd'

I did the test on a Release build of Slicer build on either 32 or 64bit and it works as expected.

Using the method described here, it should provide you with an easy mechanism to track down the issue.

From a regular "cmd.exe" terminal, you could also run the following test:

ctest -C Debug -R SlicerPythonSimpleNUMPYTest -V

or

ctest -C Release -R SlicerPythonSimpleNUMPYTest -V

Hth
Jc

See http://massmail.spl.harvard.edu/public-archives/slicer-devel/2012/008170.html

From Jc email - March 27:

Hi Folks,

With the help of Daniel and Steve :), it resumes the problem is probably related to the fact manifest associated with the pyd module are not properly managed.

The following topic [1] contains the required change.

To apply locally the change in your source tree, you have two options:

1) If you are using git-svn:
cd Slicer
git stash # In case you have local changes
git remote add jcfr git://github.com/jcfr/Slicer.git
git checkout -b patch-python-msvc9compiler jcfr/patch-python-msvc9compiler

2) if you are using SVN:
Download the patch available here [2]
Apply it using for example tortoiseSVN

In both case, before configuring and rebuilding, make sure to delete from Slicer-Superbuild:

• NUMPY*
• python*
• CLAPACK*
• tcl*

Let us know how it goes, if you are able to successfully:
import numpy.core.numeric
numpy.core.numeric.multiarray

I will then integrate the change to master.

Hth
Jc

[1] https://github.com/jcfr/Slicer/compare/patch-python-msvc9compiler
[2] http://lcorneliussen.de/utils/git-svn/github.com/jcfr/Slicer/tree/patch-python-msvc9compiler

See http://massmail.spl.harvard.edu/public-archives/slicer-devel/2012/008173.html

Steve> When you have chance, you could try the solution described in the the following note: http://na-mic.org/Mantis/view.php?id=1747#c3953

I did a rebuild of the indicated package after merting jcfrSlicer/patch-python-msvc9compiler and was able to import numpy with no problem :) Let's merge this into the trunk. Thanks!

Extra. Thanks for checking :) ((also thanks github))

Integrated to Slicer trunk - See http://viewvc.slicer.org/viewvc.cgi/Slicer4?view=revision&revision=20029

Dear issue reporter,

Good news :) Slicer developers SOLVED the problem you reported - YOU now need to VERIFY and CLOSE this issue.

Fix committed to 1999-expose-dcmtk-dir-to-built-in-module branch.