[Insight-users] Re: ITK Wrappers and gcc_xml

[Zachary Pincus]

>> Out of curiosity why are you trying to wrap such a low-level
>> interface?
>
> ... and a good question.  Actually, I don't need it and neither does
> Zach, but Gaetan has indicated that it's handy for for example
> hard-coding the number of processes on hyperthreaded CPUs.
>
> I'll probably disable it in my WrapITK checkout.
>
> Zach and Gaetan, should this be configurable in the CMake config?
> Disabling troublesome wrappings (for windows), such as MultiThreader
> and ImageBoundaryCondition?

So the deal is that there is some high-level stuff in  
itk::MultiThreader, like controlling the number of threads allowed.  
Right now, when gcc_xml is unleashed on that file, it finds some  
really low-level stuff too and tries to wrap that, and fails on  
windows machines.

The solution that I am working on is to feed gcc_xml a hacked-up  
MultiThreader header on windows builds, so that windows users will  
get control over the high-level stuff, but not have the low-level  
problems.

Note also that the link issues caused by similar gcc_xml + cswig  
interaction issues  are not for "low-level" things at all, but  
NumericTraits::Zero and One. Right now we're not trying to wrap the  
NumericTraits, but it wouldn't be a bad idea -- except for the link  
problem. So again, the solution will be to provide gcc_xml on windows  
with hacked-up headers for ImageBoundaryCondition (which eventually  
includes NumericTraits...).

Charl, your solution of removing MultiThreader and BoundaryCondition  
will work fine for the time being. I'd rather try to get slightly- 
better workarounds in the main WrapITK package, though. Hopefully by  
tues or wednesday I can look into that.

Zach