Re: itk-4.5 and python
On Wed, Feb 26, 2014 at 6:06 PM, Gert Wollny <gw.fossdev@gmail.com> wrote:
> To document a bit where I stand with this endeavor: so far I hit three
> problems,
>
> the first is a cmake bug:
>
> http://public.kitware.com/Bug/view.php?id=13794,
>
> Indicator: all python test fail.
> The workaround is to set the preferred python version on the cmake
> command line.
>
This is a long standing issue that effects many projects. I've pushed
a patch to stage based on the comments in the bug report and will do
my best to get it resolved.
> The second one seems to be a problem with g++. With aggressive
> optimization flags it issues code that wants to read data from an
> unaligned address with an xmms operation that requires aligned data. The
> indicator was that the itkBoxSmoothFilterTest 3 and 10 failed. If I get
> around to create a minimal test case I might even be able to properly
> report a bug.
> Luckily, with the default Debian build flags this bug is not triggered.
>
> The final one is a real show stopper:
>
> https://issues.itk.org/jira/browse/ITK-3241
>
> i.e. some optimizers are not properly wrapped, and hence some python
> tests fail.
There have been recent patches that address some of these warnings. I
backported them to the "release" branch, which will become 4.5.2.
The python wrapping is in flux. We could do a short online hackathon
to work on remaining issues if you are interested.
Another library to consider packaging is SimpleITK [1], which has very
good python wrapping.
Thanks,
Matt
[1] http://www.simpleitk.org/
Reply to: