# GROMACS 2018.4 release notes¶

This version was released on November 12, 2018. These release notes document the changes that have taken place in GROMACS since version 2018.3, to fix known issues. It also incorporates all fixes made in version 2016.5 and earlier, which you can find described in the Release notes.

## Fixes where mdrun could behave incorrectly¶

### Correct PME forces with free energy without perturbed charges/LJ¶

With free-energies calculations with lambda not set to zero and no actual perturbed charges or atom types for Lennard-Jones, the Coulomb or LJ PME mesh forces would be scaled with lambda. Note that this bug did not affect the, usual, setup where charges or atom types are actually perturbed.

Issue 2640

### Add constraint contribution to foreign Hamiltonian differences¶

The contribution of perturbed constraints was missing from the foreign Hamiltonian values. This is important for free energy calculations, such as BAR.

Issue 2703

### Add mass contribution to foreign Hamiltonian differences¶

For free energy calculations with perturbed masses, the kinetic energy contribution was missing from the foreign Hamiltonian values.

Issue 2703

### Work around bugs with expanded ensemble runs¶

With expanded ensemble runs, the energies would be outdated or zero with the velocity Verlet integrator with nstcalcenergy>1 or with other integrators when nstexpanded was not a multiple of nstcalcenergy. In these cases mdrun now sets nstcalcenergy to 1.

### Checkpoint continuations require suitable .tpr files¶

The step of a checkpoint file used for an mdrun restart must now be less than the number of steps in the .tpr. Formerly, the step in the checkpoint could be any number, and mdrun -nsteps could be used to get a particular result, but the use of that option is already deprecated. Use gmx grompp or gmx convert-tpr to make a .tpr file that expresses the intent.

Issue 2717

## Fixes for gmx tools¶

### Fix mindist output file checks¶

mindist would not check if the output file needed to print residue names and residue contacts over time was actually defined, leading to errors with empty file name strings.

Issue 2653

### Fix gmx helix segmentation faults¶

The .tpr file is now read correctly, and the helix analysis correctly handles selections that include proline residues.

Issue 2701

### Fix bug in entropy calculation in gmx anaeig¶

When gmx anaeig received an inconsistent number of atoms and eigenvectors (fewer eigenvectors than three times the number of atoms) the entropy calculations would use uninitialized values.

Issue 2668

## Miscellaneous¶

Fixed an issue where the log file could sometimes report an incorrect initial dynamic load balancing state

Issue 2631

### Fix Bromine parameters in amber forcefield files¶

The forcefield entries for Bromine and Iron were missing the actual values to define sigma and epsilon. The proper values have been included from parm99.dat for Bromine. As Iron has no corresponding parameters, the entry has been removed.

Issue 2711

## Made normal-mode analysis work for more than one molecule¶

Fixed an issue where normal mode analysis would only consider the first copy of each molecule in a system. Also fixed issues with vsites or shells in normal modes.

Issue 2720

### Disallow rerun using same filename as output file¶

When using identical filenames for -rerun a cryptic error was thrown because the same file would be used for reading and writing. Now mdrun will give a helpful error message to get around this.

Issue 2634

### Fix issue when building GROMACS without TNG¶

Some compiler errors have been resolved that could show when building GROMACS without TNG support enabled.