[lammps-users] Segmentation fault when using compute group/group and hybrid pair style

Dear all,

I’m trying to run a simulation using the hybrid pair style using AIREBO and an LJ + Coulomb potential.
I’m using the latest LAMMPS version (29 Oct 2020) compiled with the options:

When I run the attached input with the “compute group/group” line commented, LAMMPS runs just fine.

However, if I uncomment that line I get a segfault.

Am I missing something? I looked into the documentation of both hybrid pair style and compute group/group but did not see any warnings concerning the compatibility of both.

With this same compilation, I can use compute group/group with just lj/cut/coul/long just fine.



segfault.tar.xz (48.3 KB)

This is a complex and unusual configuration. You can see that LAMMPS is creating 5 neighbor list requests to process it.
So it is quite possible that you are running into a bug.
But figuring something like that out requires the hand of the “Grand Master of ye olde neighbor list code” (cc’d).


If you don’t get a response within a few days, please consider submitting an issue with your reproducer example attached to the LAMMPS project issue tracker on github.

mailing list emails have a tendency to drop from people’s radar after a while (and the mailing list has a bit of traffic, too). github issues persist until they are explicitly closed.
mind you that doesn’t include any promises to resolve this in a timely fashion. we do what we can in the amount of time available, but there is just too much to do and too little time and too few people. :frowning:


Dear Axel,

Thank you for your reply.

I hope Steve can give some input, but in the meantime, I’ll create an issue anyways, so it’s easier for you to track the bug down.

Don’t worry, I know maintaining a code as complex as LAMMPS is not an easy task. I maintain some software too (although MUCH smaller), and I know how hard it is too keep track of everything and manage our times.

Btw, thank you for doing such a great job with LAMMPS!