code_saturne User's Forum

Hi
 
0)
I had an working MPI installation of Code Saturne, which i was using to test a few examples.

1)
When playing with turbulence model, the program started to crash and/or MPI started to crashed after a few iterations. I was "playing" [usuall hypothesis not really respected] so i thought it was a modelling problem.
 
But when i reverted back to "safe calculation", MPI would always crash with 2 processors after a few iterations and then wouldn't even start.
 
2)
Now, even the "Tjunction Tutorial" doesn't work any more on my computer using 1 single core and previously working xml file.
 
3)
I'm attaching the error file hoping you can help with this issue.
 
Thanks in advance

ps: i found it weird that now, CS seems to stumble on thermal conductivity?

:| humm, your file of parameters (i.e. the xml file) is corrupted: the information about the label of  "thermal_conductivity" is contained twice.
Yust start a new file of parameters from scratch and delete this one.
 
However, may be you find a bug of the GUI. If you have this error again, please try to post the sequence of commands that create this error.

Could you attach your corrupted xml file please ? Thanks in advance.

Here is the xml file from the temporary folder
I will also check my other "bad case"
 
Thanks for the quick answer

OK, I think you find a bug :
Several attributes found: 2
The first one is Thermal_conductivity
Xpath: //property[@name='thermal_conductivity']/@label

means that the attribute "label" for the property "thermal_conductivity" is found twice. And in the attached xml file, there are :
<property label="Thermal_conductivity" name="thermal_conductivity" support="boundary"/>
and
<property choice="user_law" label="ThermalCond" name="thermal_conductivity">
A quick reparation is to clean the xml file by hand: just delete everything between the markups "radiative_transfer" (you don't activate the thermal radiative transfer in you case).
The right correction would be a new Xpath query with:
//property[@name='thermal_conductivity' and not(@support)]/@label

but I need time to implement and test it. I will post here a patch ASAP.
 
Anyway thanks for the feedback ;)

Laurent, could you try this patch? Just ungzip/untar the archive in your SRC directory. Keep your file of parameters unchanged.

Hum

Are you sure i "just ungzip/untar" while keeping my file as the original one i posted?

I went to /home/malodpan/saturne-2.0-rc1/installer/ncs-2.0.0-rc1/src/base/, gunzipped the file, changed the name to patch_ray_guy_100422.tar, detarred it but i still have the issue

Should i clean and recompile something?
 
Regards

Sorry I am not enough clear: ungzip/untar in the SRC directory of your case (where there are DATA SRC SCRIPTS RESU)!
In fact you have modified your Code_Saturne sources: if you want to take into account this modification you have to recompile the whole sources.
If you don't want to change the standard sources, just use the SRC directory that allows to modified Code_Saturne locally, only for the concerning case.

Ok!
1) " you have modified your Code_Saturne sources: if you want to take into account this modification you have to recompile the whole sources" 
=That's what i thought!

2) "the SRC directory that allows to modified Code_Saturne locally, only for the concerning case"
Works perfectly.

3) Another thing, not linked i guess: my OpenMPI cases are working again.
I don't know what happened regarding this but well, if it happens again i will try to really monitor that.


Thanks for your help!

:D
you're welcome.  If you compute some pretty results, you should show them in the forum "Flow modeling examples" ;)
 
PS: I rename the thread with a convenient label