You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
GROMACS: periodic molecules (bonded across the cell boundary), and a first-run walkthrough #262
1. periodic-molecules. The .mdp ECCE generates has no periodic-molecules setting, and the editors offer none. Infinite periodic systems (clays, other minerals, MOFs, crystals; the MINFF clay set uses periodic-molecules = yes) have bonds through the periodic boundary, so they need it, or mdrun fails or computes them wrongly. Add it to the GROMACS editors (on by default when the topology has bonds longer than half a box vector, or when the structure is marked periodic), and write it into the .mdp.
2. Help: a GROMACS walkthrough, like qe-first-calculation: a small system (a water box or a small solvated molecule) through minimisation → NVT → NPT → MD in ECCE. It should cover:
Found 2026-10-11 while rendering a real MINFF clay system (3,371 atoms): the Builder cannot open a GROMACS .gro file (FragUtil::load reads XYZ, MVM, CAR and PDB only, although mimetypes lists .gro), so the box is lost: no cell outline, a clay layer straddling the boundary shows as two half layers, and bonds cut by the boundary are simply missing. Also .xtc/.trr trajectories cannot be played (#180). Both belong to the GROMACS round: a .gro reader with the box as the cell, periodic bonds, and trajectory playback.
Merged test build (wip/test-912 8b056c5), 2026-10-15. Windows (NerdEgg): .gro is listed in Add Structure from File; em.gro (10,236 atoms) loads with the cell, and rotation is workable with no 'not responding'. macOS: .gro is listed, but ~/Downloads is unreadable to ECCE (macOS folder protection, see #344). Linux check pending.
Bug: after attaching em.gro + processed.top to a GROMACS Optimize task, the task's structure shows atoms only: no bonds and no hydrogens. The Builder's own .gro import keeps both, so the loss is in the attach path (GromacsInputsPanel: Fragment::restoreGRO, then MdTask::fragment storing it). To trace.
Activity
ohlincha commentedon Oct 11, 2026
Found 2026-10-11 while rendering a real MINFF clay system (3,371 atoms): the Builder cannot open a GROMACS .gro file (FragUtil::load reads XYZ, MVM, CAR and PDB only, although mimetypes lists .gro), so the box is lost: no cell outline, a clay layer straddling the boundary shows as two half layers, and bonds cut by the boundary are simply missing. Also .xtc/.trr trajectories cannot be played (#180). Both belong to the GROMACS round: a .gro reader with the box as the cell, periodic bonds, and trajectory playback.
Builder opens and imports .gro; groToPdb writes the triclinic cell (#262
.gro reader: atom names starting with O or H are oxygen or hydrogen (#…
docs/claude: groToPdb now writes the triclinic cell (#262)
Builder: an imported structure that brings a cell opens the Periodic …
GROMACS: attaching a structure with a box ticks Periodic molecules wh…
GROMACS: an attached .gro is read with its box, so the task's structu…
ohlincha commentedon Oct 15, 2026
Merged test build (wip/test-912 8b056c5), 2026-10-15. Windows (NerdEgg): .gro is listed in Add Structure from File; em.gro (10,236 atoms) loads with the cell, and rotation is workable with no 'not responding'. macOS: .gro is listed, but ~/Downloads is unreadable to ECCE (macOS folder protection, see #344). Linux check pending.
ohlincha commentedon Oct 15, 2026
Linux (niobium, Wayland), merged build: