OpenMM hangs and uses all system memory, tug atom mouse mode

[Tom Goddard]

Using the tug mouse mode on Mac OS 10.11 and 10.12 frequently at the first atom tug attempt ChimeraX hangs and very quickly uses all system memory (about 2 Gbytes / second). For example, opening 1gcn and adding atoms

open 1gcn
addmissingatoms #1

then use tug mouse mode. I believe this happens when their is very bad geometry. The addmissingatoms uses PDBFixer which adds atrocious geometry. I've attached a file where it put GLN 3 HG2 and HG3 atoms exactly on top of each other, which is probably the cause of the hang for this case. Each run of PDBFixer generates different results -- some randomization is involved.

This bug is particularly serious as it disables the Mac if ChimeraX is not force quit within 30 seconds or so, after which it is very difficult to force quit because the machine becomes unresponsive.

[Tom Goddard]

PDBFixer added hydrogens with identical positions

[Tom Goddard]

The attached PDBFixer output PDB version of 1GCN has hydrogens exactly on top of each other.

ATOM 41 HG2 GLN A 3 44.065 27.765 6.376 1.00 0.00 H
ATOM 42 HG3 GLN A 3 44.065 27.765 6.376 1.00 0.00 H

Residue MET 27 also has hydrogens nearly on top of each other

ATOM 436 HG2 MET A 27 5.956 16.482 14.457 1.00 0.00 H
ATOM 437 HG3 MET A 27 5.964 16.481 14.463 1.00 0.00 H

Other HG2/HG3 pairs are well separated.

[Tom Goddard]

Hangs occur when simulation goes unstable in OpenMM. Tristan adapted code using Langevin integrator and chedks for big coordinate changes which then insert minimization and the hangs have not been encountered since those changes.