[gmx-users] angles without bonds - SegFault

[Berk Hess]

>From: Marcin Wojdyr <wojdyr at unipress.waw.pl>
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>Subject: RE: [gmx-users] angles without bonds - SegFault
>Date: Fri, 3 Dec 2004 15:39:25 +0100 (CET)
>
>On Fri, 3 Dec 2004, Berk Hess wrote:
>
> > This is by design.
> > We assume that all normal 'chemical' interations are only through bonded
> > atoms.
> > This saves some (but maybe a neglegible amount of) time in the code.
> > If you want angles without bonds you need to add the IF_GRAPH flag to
>
>Thanks, it seems to work (it starts simulation).
>
>But the program still behaves differently when there are no bonds
>and when there are zeroed bonds. Unfortunatelly I don't have any good
>topology for testing (I still have problems with my ZnS) so I tested it on
>methanol from tutorial with constraints=none.
>
>If bonds are set to zeros, mdrun just segfaults, but if there are no
>bonds, the output says it is an internal bug:
>
>[...]
>Step 2  Warning: pressure scaling more than 1%, mu: 179.299 179.299 179.299
>-------------------------------------------------------
>Program mdrun, VERSION 3.2.90_beta_20040828
>Source code file: nsgrid.c, line: 220
>
>Fatal error:
>Number of grid cells is zero. This should never happen, and
>is either due to an internal Gromacs bug or a compiler error.
>
>-------------------------------------------------------
>
>Is the second error caused only by unphysical forces and will it
>work with well-constructed topology (without bonds)?

That is what I would guess.
Methanol without bonds will rapidly collaps and cause instabilities.
So the error message is actually incorrect. This can happen and
is not due to an internal Gromacs bug of a compiler error.

Berk.

_________________________________________________________________
Play online games with your friends with MSN Messenger 
http://messenger.msn.nl/