[Gmsh] gmsh 2.0.8 aborts or segfaults

Sat Dec 15 13:22:26 CET 2007

Werner Scholz a écrit :
> Hi Christophe!
>
> This is a follow-up on the previous discussion (see below) but I have a
> new problem.
>
> It seems that I was able to fix problem (1) and now the points are
> sufficiently far apart to keep the mesh algorithms happy. (I think I
> achieved this just by scaling the whole model.)
>

Hello, you should not use netgen here. There are too much volumes.

With the inhouse 3D mesh generator, gmsh is
able to produce the 3D mesh in about 5 seconds on my laptop.
In fact, gmsh generate the mesh for all volumes at once and 
re-distribute elements
in the volumes

I use the nightly version (as a developer),

My command line is

./bin/gmsh ~/Desktop/media650.geo -3 -saveall -optimize

I got rid of the line specifing the 3D algo in your input file.

JFR

> However, now in some cases gmsh does not want to save the mesh even
> though the meshing itself was successful! Instead gmsh throws the error
> message
>
> "Unable to open file 'media651b.msh'".
>
> I think I made sure there are no problems with disk space, read
> protected files of the same name, or other trivial reasons for this 
> problem.
>
> Please find attached two examples (geo files gz compressed):
>
> media650.geo        gmsh geo file which works fine with "gmsh -3"
> media650.geo.png    snapshot of the model
>
> media651b.geo        gmsh geo file which triggers the problem
> media651b.geo.png    snapshot of the model
>
> media800.geo        larger gmsh geo file which triggers the problem
>
> There is only a tiny difference between the media650 and media651 
> models in the upper left corner: model651b has one "cell" more than 
> model650.
>
> Meshing with "gmsh -3" and NETGEN works fine in both cases.
>
> Just with media651b.geo gmsh does not save the mesh properly (both from
> the command line and in the GUI). To my surprise, my self-compiled 
> version of gmsh saves the mesh of media651b correctly (it uses netgen 
> too, but the 3D meshing performs a different number of swaps). 
> However, both the gmsh Linux binary from the gmsh website as well as 
> my self-compiled version fail with media800.geo with the above 
> mentioned error message.
>
> Any help would be greatly appreciated.
>
> Thanks,
>
> Werner
>
>
> Christophe Geuzaine wrote:
>>
>> Hi Werner - After a quick investigation it seems that the problem is
>> twofold:
>>
>> (1) some points in your file are very close to each other (almost
>> identical): this leads to the creation of extremely "thin" surfaces in
>> the extrusion, which in turns makes the 2D and 3D mesh algorithms 
>> unhappy
>>
>> (2) your file also seems to trigger a weird bug in our code (the "Weird
>> point" message)
>>
>> Could you regenerate a .geo file that fixes (1)? Then we could
>> investigate (2) in more detail.
>>
>> Christophe
>>
>> magpar wrote:
>>> Hello again,
>>>
>>> Just a quick follow-up on the problem reported previously:
>>>
>>> The second geo file attached to the previous email, which shows the 
>>> problems, has a bunch of distorted volumes, which show up when 
>>> rendering the geo file in gmsh. Usually my scripts remove these 
>>> volumes and they do not appear to cause the segfault problems.
>>>
>>> Attached another example, which causes gmsh to segfault. This time 
>>> without the distorted volumes.
>>>
>>> When opening this geo file with gmsh in the graphical user 
>>> interface, gmsh throws these errors:
>>>
>>> gmsh media_p03.geo
>>>
>>> Info    : Reading 'media_p03.geo'
>>> Error   : Weird point 554 in Coherence
>>> Error   : Weird point 554 in Coherence
>>> Error   : Weird point 510 in Coherence
>>> Error   : Weird point 510 in Coherence
>>> Error   : Weird point 510 in Coherence
>>> [...]
>>>
>>> However, gmsh does display a reasonable geometry.
>>>
>>> Meshing leads to the segfault:
>>>
>>> gmsh -3 media_p03.geo
>>> (Mesh.Algorithm3D=4;)
>>>
>>> Info    : 16 points, 23 elements
>>> Info    : Meshing volume 23 (Netgen)
>>> Fatal   : Segmentation violation (invalid memory reference)
>>>
>>> gmsh -3 media_p03.geo
>>> (Mesh.Algorithm3D=1;)
>>>
>>> Info    : Meshing volume 473 (Delaunay)
>>> Info    : Meshing volume 474 (Delaunay)
>>> Fatal   : Segmentation violation (invalid memory reference)
>>>
>>>
>>> Thanks in advance for any help.
>>>
>>>
>>> Werner
>>>
>>>
>>> ------------------------------------------------------------------------ 
>>>
>>>
>>> _______________________________________________
>>> gmsh mailing list
>>> gmsh at geuz.org
>>> http://www.geuz.org/mailman/listinfo/gmsh
>>
>>
>
>
>
> ------------------------------------------------------------------------
>

-------------- next part --------------
A non-text attachment was scrubbed...
Name: media650b.png
Type: image/png
Size: 76048 bytes
Desc: not available
URL: <http://www.geuz.org/pipermail/gmsh/attachments/20071215/1836057d/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: media650.png
Type: image/png
Size: 69106 bytes
Desc: not available
URL: <http://www.geuz.org/pipermail/gmsh/attachments/20071215/1836057d/attachment-0001.png>