this partial specialization was only used for testing, so it can be removed
The problem comes from the std::vector of mutex(s), which can't be copied
using the std::vector operator=
Fix the example examples/BGL_triangulation_2/emst.cpp:
the filtered (finite) triangulation was not used.
Tested in CGAL-4.5.1-Ic-167.
Approved by the Release Manager.
With MinGW, __declspec is also implemented.
Reported in the CGAL-discuss mailing list:
Date: Wed, 26 Nov 2014 14:59:46 +0000
From: xantares 09 <xantares09@hotmail.com>
To: "cgal-discuss@inria.fr" <cgal-discuss@inria.fr>
Subject: [cgal-discuss] MinGW issues
That is a follow-up to that commit:
| commit eefe2012c0
| Author: Laurent Rineau <laurent.rineau@cgal.org>
| Date: Thu Oct 2 16:59:19 2014 +0200
|
| Bug-fix in Mesh_3 with time stamps
|
| The implementation of the class template Protect_edges_sizing_field was
| not correct when time stamps are used for Vertex_handle. The issue is
| that the time stamp of a vertex change when one do:
| tr.remove(v);
| v = tr.insert(point);
| The vertex pointed by 'v' is re-used by the TDS, but the time stamp of
| the vertex does change.
|
The fix was not complete.
The changes the interface ID declared by
demo/Mesh_3/include/CGAL_demo/Plugin_interface.h, so that the ID of
plugin interface for the Polyedron demo and the Mesh_3 are
different. That will allow to have plugins for both demos in the same
directory.
it collects edges incident to the internal facets of the conflict zone,
that are restricted to the surface.
For each such edge :
- if it appears once, we keep it
- if it appears twice, we don't consider it
- if it appears more than twice, we are in a non-manifold configuration,
and the move (pump) is canceled
Fix a bug in Mesh_3, when timestamps are used to compare Vertex_handle
and Cell_handle. The bug was in the class
Protect_edges_sizing_field. Once a vertex is removed from a
triangulation, the timestamp of the Vertex can change. That means that
it invalidates any future use of the original Vertex_handle.
Bug-fix for CGAL-4.5, approved by the Release Manager.
fix the use of "revert_move" that was producing holes on the mesh surface
infinite cells don't need to be backuped, but the data of facets that belong to
the convex hull should be restored (from inside the mesh)
The implementation of the class template Protect_edges_sizing_field was
not correct when time stamps are used for Vertex_handle. The issue is
that the time stamp of a vertex change when one do:
tr.remove(v);
v = tr.insert(point);
The vertex pointed by 'v' is re-used by the TDS, but the time stamp of
the vertex does change.
we still don't backup infinite cells because the infinite vertex does not have
a proper vertex ID (i.e. meshing_info())
However the facet data should be restored in Delaunay-restricted facets that
belong to an infinite cell (i.e. mirror facets of restricted facets that
belong to the convex hull)
Laurent Rineau
Jane Tournois
Sébastien Loriot
Andreas Fabri
Clement Jamin