CGAL users discussion list

["Laurent Rineau (GeometryFactory)" <>]

Le mercredi 16 février 2011 13:07:05, tochaf a écrit :
> Hy
> 
> I had installed libcgal 3.5.1-1 from synaptic on my ubuntu 10.4.
> 
> Now, i able to use some functions of CGAL, as instance of:
> 
> #include <CGAL/Constrained_Delaunay_triangulation_2.h>
> #include <CGAL/Triangulation_hierarchy_2.h>
> #include <CGAL/Constrained_triangulation_plus_2.h>
> 
> However, when i try to use:
> 
> #include <CGAL/Exact_predicates_inexact_constructions_kernel.h>
> 
> I lead to:
> 
> 
> /usr/include/boost/operators.hpp:129: error: template instantiation depth
> exceeds maximum of 50 (use -ftemplate-depth-NN to increase the maximum)
> instantiating ‘struct
> boost::equality_comparable2<CGAL::Quotient<CGAL::MP_Float>, double,
> boost::detail::empty_base<CGAL::Quotient<CGAL::MP_Float> > >’
> /usr/include/boost/operators.hpp:129:   instantiated from
> ‘boost::less_than_comparable2<CGAL::Quotient<CGAL::MP_Float>, double,
> boost::equality_comparable2<CGAL::Quotient<CGAL::MP_Float>, double,
> boost::detail::empty_base<CGAL::Quotient<CGAL::MP_Float> > > >’
> /usr/include/boost/operators.hpp:409:   instantiated from
> ‘boost::totally_ordered2<CGAL::Quotient<CGAL::MP_Float>, double,
> boost::detail::empty_base<CGAL::Quotient<CGAL::MP_Float> > >’
> 
> 
> and so on, a lot of errors.
> 
>  I use gcc as "compilator" in my eclipse projects.
> 
> Thus, is it a gmp issue? Why
> Exact_predicates_inexact_constructions_kernel.h "doesnt work"?

This is not a GMP issue.

It seems your Eclipse project uses the compilation flag 
"-ftemplate-depth-50". 
Modify the Eclipse project to remove it, or use "-ftemplate-depth-1000" 
instead. You also absolutely need the compilation flag "-frounding-math" if 
not present, and absolutelu NOT "-ffast-math" (remove it or use 
-fno-fast-math 
instead).

Probably the CGAL::Exact_predicates_inexact_constructions_kernel needs more 
than 50 levels of templates. I am not surprised because this kernel falls 
back 
to an exact arithmetic using CGAL::Quotient<CGAL::MP_Float>, and that number 
types uses the Boost.operators library (from www.boost.org), that needs a lot 
of templates levels.

Note that it seems your CGAL installation is configured not to use GMP, but 
CGAL::MP_Float instead for exact arithmetic. It would be better to configure 
CGAL with GMP, because GMP is faster than the internal class CGAL::MP_Float. 
You should add "-DCGAL_USE_GMP" and "-DCGAL_USE_MPFR" to your compilation 
flags. Those flags you be in /usr/include/CGAL/compiler_config.h, and I 
wonder 
why they are not defined...

-- 
Laurent Rineau, PhD
R&D Engineer at GeometryFactory           http://www.geometryfactory.com/
Release Manager of the CGAL Project       http://www.cgal.org/