MantisBT - Rosetta
View Issue Details
0000278Rosetta[All Projects] Input Handlingpublic2013-05-31 10:032014-07-19 10:55
docking in particular, anything with fullatom/centroid mode and constraints
-cst_file file_with_atompairconstraints
0000278: AtomPairConstraint does not check existence of atoms (?)
This is a bug I've run into in old versions of Rosetta (3.4 and earlier). A casual reading of trunk suggests this has been fixed but we need to check it.

AtomPairConstraint does not appear to check if atoms actually exist before attempting to apply constraints. If you pass fullatom constraints into centroid mode, where the sidechain atoms do not exist, the constraint functions look at garbage atom positions and return garbage constraint scores. Occasional crashes occur from segfaults, but usually it looks at allocated but inappropriate memory.

Here's a 3.4 example. [^]

I have not yet attempted to reproduce this in trunk.
No tags attached.
Issue History
2013-05-31 10:03smlewisNew Issue
2013-08-15 10:02amw579Note Added: 0000313
2014-07-19 10:54rmorettiFixed in SVN Version => ???
2014-07-19 10:54rmorettiNote Added: 0000350
2014-07-19 10:54rmorettiStatusnew => resolved
2014-07-19 10:54rmorettiResolutionopen => fixed
2014-07-19 10:54rmorettiAssigned To => rmoretti

2013-08-15 10:02   
OK, do you mean that this _only_ happens when the reason the atoms don't exist is because of centroid mode? Or are you concerned that this is a general failure mode when atoms don't exist?

It's definitely not true in the general case--just weeks ago I was dinged for constraining nonexistent atoms--but maybe it's somehow a specific-to-centroid-mode problem.
2014-07-19 10:54   
It's definitely been fixed by 2013wk52. This has been tested with centroid poses & sidechain constraints.

AtomPairConstraint used core::pose::named_atom_id_to_atom_id(), which checks for existence and raises EXCN_AtomNotFound if the atom is not found.