[Gmsh] Scale dependent FATAL error, multiple times in surface mesh

[Christophe Geuzaine]

Anthony Costa wrote:
> Hi All,
> 
> My geo file is as follows:
> lc = 0.000001;
> 
> ctm = 0.001;
> 
> negx = 0.495*ctm;
> negy = 0.48*ctm;
> negz = -0.008*ctm;
> posx = 0.545*ctm;
> posy = 0.496*ctm;
> posz = 0.030*ctm;
> 
> Point(1) = {negx, negy, negz, lc};
> Point(2) = {posx, negy, negz, lc};
> Point(3) = {posx, posy, negz, lc};
> Point(4) = {negx, posy, negz, lc};
> Point(5) = {negx, negy, posz, lc};
> Point(6) = {posx, negy, posz, lc};
> Point(7) = {posx, posy, posz, lc};
> Point(8) = {negx, posy, posz, lc};
> Line(1) = {4,3};
> Line(2) = {3,7};
> Line(3) = {7,8};
> Line(4) = {8,4};
> Line(5) = {4,1};
> Line(6) = {1,5};
> Line(7) = {5,8};
> Line(8) = {5,6};
> Line(9) = {6,2};
> Line(10) = {2,1};
> Line(11) = {2,3};
> Line(12) = {7,6};
> Line Loop(13) = {3,4,1,2};
> Plane Surface(14) = {13};
> Line Loop(15) = {5,6,7,4};
> Plane Surface(16) = {15};
> Line Loop(17) = {10,6,8,9};
> Plane Surface(18) = {17};
> Line Loop(19) = {12,9,11,2};
> Plane Surface(20) = {19};
> Line Loop(21) = {1,-11,10,-5};
> Plane Surface(22) = {21};
> Line Loop(23) = {8,-12,3,-7};
> Plane Surface(24) = {23};
> Surface Loop(25) = {24,18,22,14,16,20};
> Volume(26) = {25};
> Physical Volume(27) = {26};
> Physical Surface(28) = {14};
> Physical Surface(29) = {22};
> Physical Surface(30) = {18};
> Physical Surface(31) = {24};
> Physical Surface(32) = {16};
> Physical Surface(33) = {20};
> 
> It's a very small (on the order of microns) box to be meshed in 3D.
> Compiling this mesh throws an error:
> Fatal   : ERROR: Edge 4 - 9 multiple times in surface mesh
> debug!
> Aborted (core dumped)
> 
> Amazingly, if I scale the the mesh to be on the order of millimetres
> and adjust the length characteristic accordingly (by changing the
> variables 'lc' to 0.001 and 'ctm' to 1), the mesh compiles just fine.
> Further, at very low characteristic length, the mesh compiles fine.
> These mesh densities are nowhere near high enough for my problem, so
> I'm quite concerned. I feel like running into a computational
> precision boundary here is unlikely ... the mesh isn't THAT dense yet!

Can you try with a recent nightly snapshot? (It works fine over here, 
even when refining the mesh by a factor of 10.)


> 
> Thanks much,
> Anthony
> 


-- 
Prof. Christophe Geuzaine
University of Liege, Electrical Engineering and Computer Science
http://www.montefiore.ulg.ac.be/~geuzaine