[gmx-developers] Re: seg fault with distance restraints - problem in graph

Michael Shirts mrshirts at stanford.edu
Thu Feb 26 05:00:36 CET 2004

So, I've investigated this a little more.

The problem appears to be that distance restraints (and bond type=6) are set
up NOT to make chemical bonds.  However, when the fshift is calculated,
(around line 330 in disre.c, version 3.1.4) it uses the molecular graph, which
does not take the distance restraints or bonds type=6 into account.

This appears to mean that distance restraints will not work between two atoms
that are otherwise unconnected -- if they are connected by SOME other series
of bonds, it should work OK (or perhaps there are other situations it will
break down as well).  If they are not connected EXCEPT by the distance
restraint (or by bond type=6), when fshift is computed, it will gad.

Any thoughts?  Is this a bug that needs to  be fixed?


On Wed, 25 Feb 2004, Michael Shirts wrote:

> I'm playing around with the distance restraint options in gromacs-3.1.4,
> trying to get a better idea of what the results are.  I set up a model system
> with just two LJ atoms, but it keeps crashing.  Included are .top, .mdp, .gro,
> and .ndx. I debugged it, and it seems to be having problems setting up the
> bonded graph -- looking at the bonded graph, it looks like-
> *g = {maxbond = 6, nnodes = 0, nbound = 0, start = 2, end = 0, nedge = 0x0,
> edge = 0x0, ishift = 0x0}
> Which does not appear to be correct.
> When mdrun reaches the line in function ta_disres
>             ivec_sub(SHIFT_IVEC(g,ai),SHIFT_IVEC(g,aj),dt);
> it gags, because g (the graph array) isn't set up correctly.
> Any suggestions?  Am I doing something wrong to get distance restraints set
> up?
> Thanks,
> Michael Shirts


More information about the gromacs.org_gmx-developers mailing list