Hi All
I am trying to run the LIGGGHTS granular pair style with an "old" LAMMPS pair style. To do this I am using the "hybrid/overlay" command to include the effects of both pair styles. My input script runs fine with the plain granular system, but as soon as I use the hybrid/overlay command I get a segmentation error as given below (the first line shows the last line that was reached in my input script):
fix wall all wall/gran/hertz/history 1 0 mesh/gran 4 bottom top inner outer
[dwain-dunn:03288] *** Process received signal ***
[dwain-dunn:03288] Signal: Segmentation fault (11)
[dwain-dunn:03288] Signal code: (128)
[dwain-dunn:03288] Failing at address: (nil)
[dwain-dunn:03288] [ 0] /lib/libpthread.so.0 [0x7f0e2dd76190]
[dwain-dunn:03288] [ 1] /home/cerecam/LIGGGHTS/src/lmp_fedora(_ZN9LAMMPS_NS23FixWallGranHertzHistory12initSubstyleEv+0x41) [0x5818db]
[dwain-dunn:03288] [ 2] /home/cerecam/LIGGGHTS/src/lmp_fedora(_ZN9LAMMPS_NS23FixWallGranHertzHistoryC1EPNS_6LAMMPSEiPPc+0x26) [0x581a38]
[dwain-dunn:03288] [ 3] /home/cerecam/LIGGGHTS/src/lmp_fedora(_ZN9LAMMPS_NS6Modify7add_fixEiPPc+0x234e) [0x5aea42]
[dwain-dunn:03288] [ 4] /home/cerecam/LIGGGHTS/src/lmp_fedora(_ZN9LAMMPS_NS5Input3fixEv+0x1d) [0x596607]
[dwain-dunn:03288] [ 5] /home/cerecam/LIGGGHTS/src/lmp_fedora(_ZN9LAMMPS_NS5Input15execute_commandEv+0x4bd) [0x597da9]
[dwain-dunn:03288] [ 6] /home/cerecam/LIGGGHTS/src/lmp_fedora(_ZN9LAMMPS_NS5Input4fileEv+0x2a6) [0x598b04]
[dwain-dunn:03288] [ 7] /home/cerecam/LIGGGHTS/src/lmp_fedora(main+0x5f) [0x59ed97]
[dwain-dunn:03288] [ 8] /lib/libc.so.6(__libc_start_main+0xfd) [0x7f0e2da15abd]
[dwain-dunn:03288] [ 9] /home/cerecam/LIGGGHTS/src/lmp_fedora [0x46d349]
[dwain-dunn:03288] *** End of error message ***
Segmentation fault
It seems to crash when I assign the wall/gran/hertz/history boundary condition to my four walls (all imported as STL files with the mesh/gran command). If I am reading the error print-out correctly, the error originates in the "initSubStyle" class in the "fix_wall_gran_hertz_history.cpp" file.
Looking in the initSubStyle class, there are two places where an error could occur. Firstly, it checks to see if gran/hertz/history was defined and returns an error if it was not. As part of this check it determines if it was initialised using the hybrid/overlay command, so I am assuming that it is compatible with hybrid/overlay. However, I am not sure if it is correctly capturing the correct pair_style via the hybrid/overlay command. The only other potential problem area is where is gets the material properties (also in initSubStyle). This should work fine if it is picking up the correct pair_style.
There are no obvious problems in the code that I can see, so any ideas would be appreciated.
Thanks
Evan
ckloss | Sat, 02/26/2011 - 01:03
Hi Evan, nice to hear from
Hi Evan,
nice to hear from you again. From the traceback I guess you are using version 1.1.10? I would recommend to check if the error persists in 1.2.6 (available via git, repository liggghtsdev). If the error persists with the new version, feel free to send me the script and I will have a look.
regards,
Christoph
evansmuts | Mon, 02/28/2011 - 13:05
Hi Christoph Yes, it has been
Hi Christoph
Yes, it has been a while, but finally I have been able to spend some time working with LIGGGHTS.
Thanks for your reply. Updating to the development version (v1.2.6) has solved the segmentation fault error. I was just updating my v1.1.10 using git pull on the normal "liggghts" repository, so I obviously wasn't getting the latest changes.
Regards
Evan
ckloss | Mon, 02/28/2011 - 19:40
Hi Evan, good to hear that
Hi Evan,
good to hear that the problem is resolved.
the actual plan is to have two repository: liggghts and liggghtsdev
The latter is the development branch, and as soon as it gets stable, the changes are merged into the main branch
This will happen soon, as soon as we have the test harness up and running on a larger scale
Cheers,
Christoph