Re: %include

On Mon, Nov 17, 2003 at 07:15:02AM -0600, David Beazley wrote:
> Kerim Borchaev writes:
>  > Hello Marcelo,
>  > 
>  > I've built the CVS version and for that example it works. But only if
>  > working directory is ./module. Otherwise it fails with:
>  > main.i:2: Unable to find 'PartA/main.i'
>  > I think it would be consistent if it would work too.
>  > And when it is implemented you could stop searching files relative to
>  > the working dir;-)
>  > 
> 
> I'm not sure I follow--in fact, I don't follow.  Do you have a
> specific example?
> 
> -- Dave

Kerim has the following directory structure for his modules:

/modules
   /PartA
      /subsidiary0.i
      /subsidiary1.i
      /module.i
   /PartB
      /subsidiary0.i
      /subsidiary1.i
      /module.i
   /toplevel.i

What he and I have been "debating" off-line has been how SWIG could be
searching for the includes. toplevel.i looks like:

toplevel.i
----------
%include "PartA/module.i"
%include "PartB/module.i"

PartA/module.i
--------------
%include "subsidiary0.i"
%include "subsidiary1.i"

What Kerim wants is for SWIG to push "-IPartA/" onto the include search
path when PartA/module.i is included, popping -IPartA/ off the search
path after PartA/module.i has been read.

I was half arguing that this behavior is assymetric from how GCC and other
compilers work (they don't do this) and that it might be confusing to the
less sophisticated user. I was never against this feature but I'm not all
that thrilled about it either.


-scooter
_______________________________________________
Swig maillist  -  Swig@xxxxxxxxxxxxxxx
http://mailman.cs.uchicago.edu/mailman/listinfo/swig