CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   SU2 (http://www.cfd-online.com/Forums/su2/)
-   -   all rans and navierstokes test cases getting seg fault (http://www.cfd-online.com/Forums/su2/124419-all-rans-navierstokes-test-cases-getting-seg-fault.html)

jentink October 4, 2013 13:22

all rans and navierstokes test cases getting seg fault
 
I recently installed version 2.0.8 and the 2.0.6 test cases.

All the test cases under rans and navierstokes die with a 'segmentation fault' before the first iteration.

These all worked fine with the base 2.0 install.

Any tips?

jentink October 4, 2013 14:40

2.0.7 works.

austin.m October 4, 2013 16:16

Hi Tom-

I've been running 2.0.8 for a day now with no issues. I've been running my own cases, but as a check I did try /TestCases/rans/naca0012 without any problem.

Can you post your config file and a screen dump?

Also, I've noticed that SU2 will seg fault if a restart is requested but file isn't correct or not available.

taxalian October 6, 2013 16:09

1 Attachment(s)
Hi,
I got similar problem with SU2 2.0.8, the code gives segmentation fault before starting with the iterations, see the attached picture for /TestCases/rans/naca0012.

The similar test cases ran fine in 2.0.7 but while writing the solution file at specified iterations, the code gives segmentation fault and stopped. Although all the test cases ran fine with the stable release of SU2.

jentink October 7, 2013 09:10

Same thing here with 2.0.7.

I can get most test cases to run, but they won't restart from a restart file written by 2.0.7.

I'm running my own case now and it goes fine as long as I restart from original solution. I adapted my grid and now seg faulting when I try to read the adapted restart file.

jentink October 7, 2013 12:07

this is what gdb is showing when I compile 2.0.8 and run the rans/rae2822 test case:

Program received signal SIGSEGV, Segmentation fault.
0x082cc8b7 in CEulerSolver::Inviscid_Forces (this=0x94ce6f0, geometry=0x88f3f70,
config=0x88b03e0) at ../src/solver_direct_mean.cpp:3505
3505 Surface_CLift_Inv[iMarker_Monitoring] = 0.0;

jentink October 7, 2013 12:47

if i turn off monitoring in the same test case, it seg faults further down. Coord[0], Coord[1], and Coord[2] are defined. Origin[] is not.

Program received signal SIGSEGV, Segmentation fault.
0x082e2cd2 in CNSSolver::Viscous_Forces (this=0x94ce6b8, geometry=0x88f3f40, config=0x88b03e0)at ../src/solver_direct_mean.cpp:7743
7743
MomentDist[iDim] = Coord[iDim] - Origin[iDim];

jentink October 7, 2013 13:23

if i hard code the moment reference length to 1.0, it works.

seg fault happens anytime you try to reference Origin[idim].

this happens whether i use the old moment reference type in the input or the new with separate x, y, and z inputs.

economon October 7, 2013 21:04

Hi,

As they are developer releases, specific details of the restart files, configuration options, and solver defaults can change between minor releases (such as V2.0.7 -> V2.0.8), and this can sometimes cause issues. It is best to use the matching test cases/config files for your current version of SU2 (there is a set of test cases specifically for V2.0.8 available for download on su2.stanford.edu).

Note that we keep the configuration files for the test cases found in SU2/TestCases/ in the repository up to date for the current working version of the code. Does trying with this version of the config files resolve the issue? If not, can you share more details about your specific cases (configuration files, etc.)?

Tom

jentink October 8, 2013 10:26

The specific issues experienced for the rans/rae2822 test case (my posts above concerning the seg fault location) were run with a git clone of the 2.0.8 release. And the configuration file from that release. No changes.

taxalian October 8, 2013 13:34

Hi Tom,
I used the concerned configuration files for the related test cases compatible with the SU2 release 2.0.7 & 2.0.8. I don't think so, that this is really causing the code to cause segmentation fault.

jentink October 8, 2013 14:43

I cleaned out all my versions and started over with a clean compile using 2.0.8.

The test cases that were seg faulting before are now working.... Sorry about that.

I think I was just doing too many things at once.

austin.m October 8, 2013 14:59

Don't feel too bad, I had that happen to me with 2.0.6... it is frustrating! In general though I've found SU2 to be getting better with each developer release, though its challenging to keep up since the changes are not documented in the tutorials yet.

Austin

economon October 8, 2013 19:29

Glad to hear that things are working now... I completely understand that things can be a bit frustrating with the developer releases, but on the positive side, it underscores the fast pace at which improvements and upgrades are being introduced into the code.

We agree that the documentation needs to be updated, and this will of course be done alongside a very important major release of SU2 V3.0 in January.

In the meantime, don't hesitate to contact us if you see weird behavior, and we'll do our best to update users of important changes in the developer releases. Thanks for working with SU2 - your feedback truly does help improve the code.

Cheers,
Tom

taxalian October 9, 2013 14:07

Hi,
As jentink mentioned, after cleaning up all the SU2 versions and installing a fresh SU2 2.0.8 distribution resolves all the issues.


All times are GMT -4. The time now is 03:26.