[Insight-users] Itk 3.6 weird compile problem : kwsys/ios and stl directories

[J.S.Wijnhout at lumc.nl]

>Just to answer your previous questions regarding the directories
>
>                       itksys/ios
>                       itksys/stl
>
>They have been in the configuration of Utilities/kwsys/CMakeLists.txt
>for a while. See the CVS annotate of this file (lines 604-606):
>http://www.itk.org/cgi->bin/viewcvs.cgi/Utilities/kwsys/CMakeLists.txt?
annotate=1.137.2.1&root=Insi>ght
>
>
>
>    What ITK release were you using before updating to ITK 3.6 ?

I was using ITK 3.4. It could be that a small change in itkPoint.h (or
included files from this) triggered the inclusion of cstddef from
itksys, but I haven't checked this.

>Regarding your second question:
>
>
>           When to use or not to use itksys ?
>
>
>It is usually a good idea to use itksys. It provides a layer
>of portable functionalities across a large number of platforms.

I thought so (especially with VS 6, STL was supposed to be a nightmare),
but for the platforms I'm using now it is probably not a big issue (VS7
and gcc 3.3).

>One thing that could help a lot is for us to install most headers in
>the same directory. In this way, you could build ITK, and then install
>it in a different directory. Currently we install ITK with a directory
>structure that is almost a copy of the build directory. However, we
>could flatten that structure and aggregate most header files in a
>smaller set of directories. This is probably a worthwhile undertaking
>for the release of ITK 3.8.

That would be a big help for us, non-CMake users, too. In fact we used
to flatten the directory structure ourselves before putting it in CVS,
but that only slowed us down when putting a new ITK version into CVS.

Thanks for your answers and clarifications.

best,
Jeroen