mdrun should not advise users of things that don't matter
At https://mailman-1.sys.kth.se/pipermail/gromacs.org_gmx-users/2016-June/106482.html the advice from mdrun upon restart that the number of nodes is different was confusing to a user, and meant that they didn't focus on the real problem.
There are many ways in which ostensibly equivalent mdrun calls will differ, so we should not differentiate a restart on a different number of ranks, or a different build of GROMACS, as something a user needs to know about. There's no case where they need to take action, and it's not likely to be an accident of concern, either. Since the restart always writes the information about mdrun (even when appending) there is no (longer) any need to say anything about version/build/build user/parallelism mismatch.
Remove warnings on checkpoint mismatch
mdrun now only warns for mismatch in minor version, build or
number of ranks used when reproducibility is requested.
Also added a separate message for not matching precision.
#3 Updated by Mark Abraham over 3 years ago
Berk Hess wrote:
Uploaded a fix. But I kept the major version mismatch warning, since we don't guarantee that to work.
Good. Now the stderr has more like
Reading checkpoint file state.cpt generated: Sun Jun 19 11:42:59 2016 Version mismatch, current program: 2016-beta2-dev-20160616-67745cd checkpoint file: VERSION 5.1.3-dev-20160617-5833198 The current GROMACS major version is not identical to the one that generated the checkpoint file. In principle GROMACS does not support continuation from checkpoints between different versions, so we advise against this. If you still want to try your luck we recommend that you use the -noappend flag to keep your output files from the two versions separate. This might also work around errors where the output fields in the energy file have changed between the different versions.
which seems fine