[gmx-users] Segmentation fault when using fourier dihedrals in v. 4.0.5.

Teemu Murtola teemu.murtola at cbr.su.se
Tue Dec 1 12:50:31 CET 2009

On Tue, Dec 1, 2009 at 11:22, Henri Ervasti <h.k.ervasti at tudelft.nl> wrote:
> I just upgraded from Gromacs 4.0.2 to 4.0.5. I wanted to re-run some
> jobs from the previous version to see if there would be any changes in
> results. I have systems with Ryckaert-Bellemans (type 3) dihedrals and
> Fourier (type 5) dihedrals. The jobs with type 3 run without a problem,
> but with type 5 they crash immediately. In the .log file are data only
> up until the reference citing recommendations.
> What might be the problem?

Fourier dihedrals (type 5) do not work correctly in 4.0.5, and I don't
think they have worked in any earlier version either (just do a
gmxdump of a tpr file with type 5 dihedrals to check). This has been
fixed in the Git version, also in the release-4-0-patches branch. It's
possible that with some compilers, the bug would result in
segmentation faults directly when reading the topology, which you may
have witnessed.


More information about the gromacs.org_gmx-users mailing list