[Insight-developers] ITK CVS HEAD, build error with Python, signal.h

[Brad King]

Charl P. Botha wrote:

> Dear developers (and especially Brad),
> 
> The watershed wrappings seem to work.  I've been stress-testing them on 
> the 1-6 branch and I'm now busy porting them to HEAD.  However, I'm 
> getting the following build error on Debian Stable, gcc 2.95.4, ITK CVS 
> HEAD:
> 
> Building wrap_ITKPyUtils.xml from 
> /home/cpbotha/DoNotBackup/build/Insight-HEAD-
> VC6/Wrapping/CSwig/Common/wrap_ITKPyUtils.xml...
> In file included from /usr/include/signal.h:358,
>                  from /usr/include/python2.2/pyfpe.h:129,
>                  from /usr/include/python2.2/Python.h:122,
>                  from 
> /home/cpbotha/DoNotBackup/build/Insight-HEAD/Wrapping/CSwi
> g/Common/itkPyCommand.h:22,
>                  from 
> /home/cpbotha/DoNotBackup/build/Insight-HEAD/Wrapping/CSwi
> g/Common/wrap_ITKPyUtils.cxx:17:
> /usr/include/bits/sigthread.h:36: error: storage class specifiers 
> invalid in
>    parameter declarations
> /usr/include/bits/sigthread.h:36: error: storage class specified for 
> parameter
>    `parameter'
> 
> Now, it seems that the compatibility header in CableSwig (pthread.h) 
> would fix this problem in sigthread.h[1], but the Python header is 
> bypassing that: Python.h -> pyfpe.h -> signal.h -> sigthread.h
> 
> It would seem that a signal.h compatibility header is also required, as 
> including signal.h (WITHOUT pthread.h) is a valid way of getting to 
> sigthread.h.
> 
> If this is so, should I add a signal.h header similar to the pthread.h 
> header to all the GCC_XML compatibility header directories, or is there 
> a better way(tm)?

Yes, please add the signal.h wrapper to to all the places that have a 
pthread.h wrapper.  Please test it by including all combinations of 
relevant headers in all orders.  We do not want two places to define 
__thread at the same time.

Thanks,
-Brad