[FLASH-USERS] Segmentation fault.

[Sean Couch]

Hi Carolina,

SegFaults don't always mean you're running out of memory, just trying to access it incorrectly.  Fortunately they are usually easy to debug.  I'd suggest compiling FLASH in debug mode (by adding '-debug' to your setup line, assuming appropriate debug flags are set in the Makefile.h for your site).  If the compiler doesn't catch anything, running the debug-compiled code usually does.  You could also use a debugger such as gdb.  If that doesn't work, I recommend going with a software called valgrind.  It will find EVERY memory error in the code (there should be none...).  It may already be installed on the cluster you're using.

Best,
Sean

--------------------------------------------------------
Sean M. Couch, Ph.D.
Hubble Fellow
Flash Center for Computational Science
Department of Astronomy & Astrophysics
The University of Chicago
5747 S Ellis Ave, Jo 315
Chicago, IL  60637
(773) 702-3899
www.flash.uchicago.edu/~smc



On Apr 25, 2013, at 4:36 PM, Carolina Pepe <carolina.pepe at gmail.com> wrote:

> Hello all,
> 
> I've been working with a 1d spherical configuration of FLASH.
> Now, I'm trying to read some tabulated values from a .dat file and change how certain variables are calculated. This reading is done in Driver_sourceTerms and Gravity_accelOneRow.
> 
> Besides this reading, no major changes have been done (just some little algebra with the read values).
> 
> However, i get this error:
> 
> mpirun noticed that process rank 0 with PID 3141 on node blackhole exited on signal 11 (Segmentation fault).
> 
> right after the first evolution step. The strangest thing of all is that it runs for certain combination of parameters (in the .par) but I get this error for some other combination.  From what I've found around, it is usually related to memory issues, but since I'm running it on a powerful cluster, this shouldn't be an issue.
> 
> Should I have extra precautions when reading a file from FLASH?
> 
> Thanks in advance for any kind of help :)
> 
> Carolina
> 
> p.s. I added an error flag to check the opening of the file, and it opens correctly. 
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20130427/f3cc31f1/attachment.htm>