Re: [K3d-development] g++ 3.0.4 troubles

[Timothy M. S. <ts...@k-...>]

Romain Behar wrote:

> I've installed a fresh Debian box and tried to compile
> K-3D using gcc 3.04:
> 
> with poly_cone.cpp, poly_cylinder, poly_grid,
> poly_sphere and poly_torus, the compilation stopped on
> the following error:
> 
> ../../boost/boost/type_traits/is_enum.hpp:91: template
> instantiation depth 
>    exceeds maximum of 50 (use -ftemplate-depth-NN to
> increase the maximum) 
>    instantiating `struct 

In the short-term, you might try to get things building with

	$ ./configure CXXFLAGS="-ftemplate-depth-100" && make

to see if that fixes things ... keep increasing the depth as-needed 
until it builds.  If we can nail-down a value that works for gcc 3.0, it 
should be harmless to set it as a default in our configuration.

The problem appears to be with the return value of k3d::add_grid(), 
which is a convenience function that creates a 2D grid of polygons 
that's useful in most of our conics.  The return value is of type

boost::tuple<boost::multi_array<point*, 2>, 
boost::multi_array<split_edge*, 3>, boost::multi_array<face*, 2> >

... which is a real mouthful.  I'd actually prefer to pass each of the 
three multi_array objects to the function by reference instead of 
returning them by value through the tuple, but multi_array can't have 
its dimensions changed after it's been constructed.  I think the 
long-term solution (something I was contemplating anyway), is to develop 
an alternative to boost::multi_array.

> BTW, the configure script didn't detect the GL/GLU
> files were not installed.

As it's been an extended period of time since I've had a machine without 
OpenGL around, these could very well have been broken for awhile (or 
since the beginning).  You might also try the previous version of 
acinclude.m4.  I recently made some changes for portability, basically I 
was using "==" with test, which is not portable.  Maybe I broke something?

Cheers,
Tim