Skip to content

[pull] develop from lammps:develop#171

Merged
pull[bot] merged 13 commits into
comphy-lab:developfrom
lammps:develop
Jul 15, 2026
Merged

[pull] develop from lammps:develop#171
pull[bot] merged 13 commits into
comphy-lab:developfrom
lammps:develop

Conversation

@pull

@pull pull Bot commented Jul 15, 2026

Copy link
Copy Markdown

See Commits and Changes for more details.


Created by pull[bot] (v2.0.0-alpha.4)

Can you help keep this open source service alive? 💖 Please sponsor : )

akohlmey added 13 commits July 10, 2026 18:17
The test generators mark new YAML files with a "generated" tag that is
to be removed once the reference data is reviewed and validated.  These
files from the recently merged test PRs had passed review and CI with
the marker still in place.
fix ffl sets scalar_flag without declaring whether its cumulative
thermostat energy is extensive, which aborts any thermo output that
accesses the scalar.  It is extensive, like the equivalent quantity
of fix langevin.

Found by the new fix-timestep-ffl.yaml test.
nlevels_respa was declared but never assigned, so any r-RESPA run with
this fix used an uninitialized level index in setup() and crashed or
corrupted the per-level force arrays.  Initialize it in init() like
its sibling fix electron/stopping does.

Found by the new fix-timestep-electron_stopping_fit.yaml test.
The base class constructor of fix wall/reflect allocates the graphics
objects for rendering walls with dump image based on the wall count of
ITS argument parser.  fix wall/reflect/stochastic does not use that
parser: it resets nwall and parses its own arguments, so the arrays
stayed sized for zero walls (a null pointer) while post_integrate()
updates them for every parsed wall.  Every run with this fix style
crashed since the wall rendering support was added.  Factor the
allocation into a helper that derived classes re-run after parsing.

Found by the new fix-timestep-wall_reflect_stochastic.yaml test.
- no_restart: stochastic fixes that do not store their RNG state in
  restart files (e.g. langevin, qtb) cannot reproduce the reference
  trajectory from a restarted run; the tag skips the restart and rmass
  legs of the test
- no_t_target: fixes that compute their target temperature internally
  (e.g. the nphug hugoniostat) cannot satisfy the check of the
  extracted t_target against the input variable
langevin (plus tally variant), ffl, qtb, baoab, viscous/nonlinear,
dt/reset, thermal/conductivity, ehex, msst, nphug, nvt/uef,
electron/stopping/fit, ti/spring, propel/self, brownian,
wall/harmonic/outside, wall/reflect/stochastic, and wall/flow.

Stochastic fixes use fixed seeds and the new no_restart tag (their RNG
state is not preserved across restarts); most also skip the r-RESPA leg
where the different force evaluation pattern changes the drawn random
stream.  msst and nphug are tagged no_restart pending a closer look at
the completeness of their restart data.  nvt/uef runs on a triclinic
box via change_box.  brownian and viscous/nonlinear run on the sphere
test system of the existing brownian/sphere test.
The fix implemented only post_force(), so under r-RESPA it was never
invoked and the propulsion force silently vanished.  Add the standard
post_force_respa()/init()/setup() pattern applying the force at the
outermost level (or the level selected with fix_modify respa).
The tag marks reference data that has not been reviewed and validated
yet and is removed manually after validation.  Previously it was only
added when a test file had no tags at all, so regenerating a file with
functional tags (like no_respa) did not mark it as unreviewed.
The TMD target geometry is the fourmol data file with all positions
randomly displaced by up to 0.4 Angstrom.  The fix computes its
distance-to-target schedule at fix creation time and stores no restart
data, so the restarted-run comparison is skipped via the no_restart
tag.  Also regenerate the propel/self reference: with its new r-RESPA
support the fix now runs under respa, but its velocity-proportional
force legitimately diverges from the verlet reference trajectory, so
the test keeps the no_respa tag.
Also clarify that the generated tag is now always added when reference
data is written and must be removed manually after validation.
Fix-timestep test additions: 21 new fix style tests (+4 bug fixes)
@pull pull Bot locked and limited conversation to collaborators Jul 15, 2026
@pull pull Bot added the ⤵️ pull label Jul 15, 2026
@pull
pull Bot merged commit e04d816 into comphy-lab:develop Jul 15, 2026
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant