[Gmsh] BUG (??) in version 4.5.6

[Christophe Geuzaine]

Dear Miguel,

The over-refinement was due to an overly strong hypothesis on the 1D mesh that was made when refactoring the code in Gmsh 4.5.4. The next development snapshot will fix this little regression.

Thanks for the report,

Christophe


> On 5 May 2020, at 15:36, miguel.ariza at artorg.unibe.ch wrote:
> 
> Dear Dr Geuzaine,
> 
> Please, find enclosed the test files.
> 
> Just as a summary of what I did: a tested the same surface and meshing conditions in two Python environments (one with GMSH 4.5.6 and another one with GMSH 4.5.3). In the case of 4.5.3, we have ~70k elements (see log-453.txt) and in the case of 4.5.6, we have ~ 1m elements (see log-456.txt). BESIDES, there is an error regarding the discretization of the curves (see lines 20-24 in log-456.txt). This error is not present in the previous version.
> 
> Hopefully, this will help you to isolate the origin of this behavior ;-D
> 
> 
> Best,
> 
> MA
> 
> -----Original Message-----
> From: Christophe Geuzaine <cgeuzaine at uliege.be> 
> Sent: Tuesday, 5 May 2020 14:03
> To: Ariza Gracia, Miguel Angel (ARTORG) <miguel.ariza at artorg.unibe.ch>
> Cc: gmsh at onelab.info
> Subject: Re: [Gmsh] BUG (??) in version 4.5.6
> 
> 
> 
>> On 5 May 2020, at 12:34, miguel.ariza at artorg.unibe.ch wrote:
>> 
>> Dear all,
>> 
>> I was using a code I developed in version 4.4 for meshing spheres with tetrahedron. I could modify the characteristic length (max. and min.) and it worked perfectly.
>> 
>> Moreover, I’ve been using this code in different incremental updates and it worked perfectly until version 4.5.3. Yesterday, I updated to version 4.5.6 and the code continues working but with a strange behavior.
>> 
>> Let’s say that I was using a characteristic length of 0.2 for meshing a sphere, in version 4.5.3 this resulted in  ~70k elements. Now, with the same characteristics, this results in 1.000.000 elements!
>> 
>> No matter what I tried, it seems that GMSH’s response was invariant:
>> 	• Activate/deactivate the optimization of the mesh (it did not change anything)
>> 	• Modify the characteristic length up to 2 or 4 (it did not change anything). Also, I debugged to be sure that these were the actual values that GMSH received
>> 	• Try to use a bias in the mesh. Nothing
>> 
>> Perhaps, I missed a major modification in the code from 4.5.3 to 4.5.6 and I need to re-code my libraries… but… Could it be a bug in the code? If so, where? In the meantime, I downgraded to version 4.5.3.
>> 
> 
> Could you send a simple example so that we can test?
> 
> Christophe
> 
>> Tested in:
>> 	• W10
>> 	• Ubuntu 18.04 in WSL (W10)
>> 	• Ubuntu 18.04
>> 
>> Best,
>> 
>> MA
>> _______________________________________________
>> gmsh mailing list
>> gmsh at onelab.info
>> http://onelab.info/mailman/listinfo/gmsh
> 
> —
> Prof. Christophe Geuzaine
> University of Liege, Electrical Engineering and Computer Science http://www.montefiore.ulg.ac.be/~geuzaine
> 
> 
> 
> <bug_gmsh456.zip>

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