Compact_mesh_cell_base_3 has a different API than Mesh_cell_base_3. Unless
one uses the default template parameters, the new API is less
convenient. We cannot deprecate Mesh_cell_base_3 so easily.
- In Mesh_3, fix the sliver perturber
- the invalidatation of caches,
- the topological check (in check_surface_mesh) was not enough.
Tested in CGAL-4.3-Ic-194
The error was triggered by the following code in CGAL/Mesh_3/C3T3_helpers.h:
template <typename CellForwardIterator>
void reset_circumcenter_cache(CellForwardIterator cells_begin,
CellForwardIterator cells_end) const
{
namespace bl = boost::lambda;
std::for_each(cells_begin, cells_end,
bl::bind(&Cell::invalidate_circumcenter, *bl::_1) );
}
See for example test results here:
CGAL-4.3-Ic-193/Polyhedron_Demo/TestReport_lrineau_i686_Linux-2.6_g++-4.1.2_CentOS-5.1.gz
The testsuite showed that our previous patch was not correct: the sliver
cache is modified by a call to C3T3_helper::min_sliver_in_c3t3_value() when
its parameter use_cache=false. For that reason, the resetting of sliver
caches was not done at the right place in update_mesh_no_topo_change().
On the paper, dleft_frac must be greater that 1.0. But with rounding
errors, it can be slightly lower. Let's try to compare it with
1.0-2*ulp(1.0).
I do not like that sort of hack but I do not know what to do best.
| commit 01ee3e94b6
| Author: Laurent Rineau <laurent.rineau@cgal.org>
| Date: Tue Jun 25 12:53:24 2013 +0200
|
| Use 'std::istream::operator!()' twice to replace 'operator bool()' (C++11 only)
|
| Fixes a compilation error from clang:
|
| include/CGAL/IO/File_binary_mesh_3.h:39:10: error: no viable conversion from 'std::ostream' (aka 'basic_ostream<char>') to 'bool'
| return os << c3t3;
| ^~~~~~~~~~
The return type of operator<< os std::ostream&, and not bool!
Fixes a compilation error from clang:
include/CGAL/IO/File_binary_mesh_3.h:39:10: error: no viable conversion from 'std::ostream' (aka 'basic_ostream<char>') to 'bool'
return os << c3t3;
^~~~~~~~~~
Successfully tested in CGAL-4.3-Ic-59
Approved by the release manager
Conflicts:
Kernel_23/doc/Kernel_23/CGAL/intersections.h
Kernel_23/doc_tex/Kernel_23_ref/intersection.tex
Kernel_d/doc/Kernel_d/Kernel_d.txt
STL_Extension/doc/STL_Extension/CGAL/iterator.h
... to be a pair< variant<XXX>, Primitive_id > rather than
pair<optional<variant<XXX> >, Primitive_id >
The rational is that the optional around the variant here is not needed
since there is one around the pair.
Laurent Rineau
Jane Tournois
Sébastien Loriot
Andreas Fabri
Michael Hemmer
Clement Jamin
iyaz