[Gmsh] possible bug in the Frontal Algo

[Johannes Wagner]

Hi Walter,
I'm glad I found your post. Just observed the same behaviour in the 
simplest case of 2d frontal algorithm, just tried to mesh a rectangle, and 
one corner triangle is missing. Just wasted an hour triple checking my 
geometry and reducing it down to just an rectangle... where the corner is 
still missing. 
Is your patch working? Perhabs you could post a diff against 2.12 source 
code? I'd like to apply it and try out if it also works for me!

cheers, Johannes



On Thursday, February 11, 2016 at 10:29:57 AM UTC+1, walter... at alice.it 
wrote:
>
> In my experience I have found that the surface triangulations generated by 
> the Frontal Algo have a high quality.
>
> Unfortunately sometimes a face triangulation is wrong because a triangle 
> is missing (see annexed picture).
> The missing triangle has always a vertex which is also a vertex of the 
> geometrical face so that the adjiacient edges belong to a couple of 
> different geometrical edges. 
>
> I have seen in your code that the option ALGO_2D_FRONTAL triggers the 
> calling of the function bowyerWatsonFrontal(..) 
> Inside of this function there is a loop where the worst triangles gets 
> destroyed and replaced by smaller tiangles upon the inserction of
> a new point. 
>
> I suspect that for that particular position (worst triangle has a vertex 
> on a face vertex) the triangle is destryed but the subsequent point 
> insertion fails.
>
> Walter
>
>
>
>
>  
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://onelab.info/pipermail/gmsh/attachments/20160414/12acefc4/attachment-0001.html>