open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

It was just the set of files (with wrong selafin format) to run a case and get the crash, so that I can try to issue an error message before the crash. If it raises permission problems for you, don't bother with this and forget it.

Regards,

Jean-Michel Hervouet

Thank you again,
I have my personal PC so I'll use this until I understand what is happening with my login. Sorry for the inconvenience!

I have corrected the file and elements should be all clockwise but I still get the same error:

[Execute@node1;16:15:56] partitioning: T2DGEO
[Execute@node1;16:15:56] +> /mnt/nfs/soft/opentelemac/v6p3r1/builds/ubugfopenmpi/bin/partel < PARTEL.PAR >> partel_T2DGEO.log
[Execute@node1;16:15:56] At line 213 of file /mnt/nfs/soft/opentelemac/v6p3r1/sources/utils/bief/elebd.f
[Execute@node1;16:15:56] Fortran runtime error: Index '-1718493944' of dimension 1 of array 'nbor' below lower bound of 1
[Execute@node1;16:15:56] ... The following command failed for the reason above
[Execute@node1;16:15:56] /mnt/nfs/soft/opentelemac/v6p3r1/builds/ubugfopenmpi/bin/partel < PARTEL.PAR >> partel_T2DGEO.log

I attached the new selafin file.





Best regards,
Maxime

PS: Jean michel, I can send you the case per email. I don't know if it will be useful given that the error is still happening but at least you'll have them to try.

Problem solved, it was simply a problem with one point of the cli file (The convention that a liquid segment must be contained between two liquid nodes wasn't respected).
The wrong orientation of elements was corrected before i could see any error thanks to you sebastien, so sorry Jean-Michel i can't help you further.
Thanks for your help!
Best regards.
Maxime JEANNIN

Hi Guys,

I faced the same problem setting up a model for the English Channel. Eventhough I used BlueKenue and all my elements are counter-clockwise I got the error of "Segmentation fault - invalid memory reference". I first had a chat with a colleague because I was trying to run it in parallel on my private machien, but it gives the same error in serial mode. I already checked and cleaned the mesh from isolated triangles.
I also set up a conceptual model and it worked fine.
Some more back tracers are listed up, but I can't find any advise about their meaning (main error message below):


READGEO1: TITLE= newSelafin

NUMBER OF ELEMENTS: 202418
NUMBER OF POINTS: 102881
MXPTEL (BIEF) : MAXIMUM NUMBER OF ELEMENTS AROUND A POINT: 9
MAXIMUM NUMBER OF POINTS AROUND A POINT: 9
CORRXY (BIEF):NO MODIFICATION OF COORDINATES

MESH: MESH ALLOCATED

*************************************
* END OF MEMORY ORGANIZATION: *
*************************************

_____________
runcode::main:
/Users/opentelemac/v6p3/EnglishChannel/SU1_1.5_3:
|runCode: Fail to run
|/Users/opentelemac/v6p3/EnglishChannel/SU1_1.5_3/SteeringEC.cas_2014-06-15-23h03min32s/out_telemac2d
|~~~~~~~~~~~~~~~~~~
|Program received signal SIGSEGV: Segmentation fault - invalid memory reference.
|
|Backtrace for this error:
|#0 0x10b9e4132
|#1 0x10b9e48ee
|#2 0x7fff92db85a9
|#3 0x10b475a96
|#4 0x10b48216a
|#5 0x10b1ab37e
|#6 0x10b1c373b
|#7 0x10b4a2392
|~~~~~~~~~~~~~~~~~~
I put my .cas, .cli, .bc2 and .slf file in the attachment.

Any help would be more than appreciated.

Cheers,
Tim

Hello Tim,
the only time I had this SIGSEGV error, it worked by rearanging the element counter clockwise. But my backtrace were different so it may be different. I can't help you further. Sorry
Still I have to say I would like to know how to use the backtrace too.
Best regards,
maxime

Hi Maxime,
yeah, I got this as a possible failure source from your conversation. But as I'm working in BlueKenue, I checked the MeshAnalyses and all elements are already counter-clockwise. Further sources of errors can be isolated triangles in the boundary file or some problems in the segmentation running t2D in parallel mode and joining the pieces back together. But I also tried to run the model in serial mode and still got the error.

I completely agree; an error shooting guide would be somewhat very useful.

But thanks anyway,
Tim

Hello Tim,

what I do in such cases is to use the stbtel utility:
simply load your mesh and save it without any processing options. Stbtel will check your mesh and list possible errors. You can access stbtel also via the Fudaa-prepro software.

In fact in your mesh - EC_1.5_3.slf - node nr. 25995 shares two outblocked areas which is not allowed in Telemac. See enclosed the picture.
As a possible solution now you could e.g. delete element nr. 51088 in order.



Hope this helps,
Clemens

Hi Clemens,

that sounds like a very useful tool. I'm not exactly sure how you access this utility, though. Do you mean by loading the mesh within python/comand window or in Bluekenue? I'm not quite sure how to use fudaa because I did everything manually till now. Is there a tool within fudaa?
I will have a closer look, but thank you very much! At least I know have problem to look after.

Cheers,
Tim

It works!!! Cheers