Selection for on_hull/hull distance isn't working properly

[jli5373]

2 issues:

1. I ran a casm select --set 'on_hull' and it selected 10 configurations, of which 2 had positive hull distances. If I then run casm select --set 'lt(hull_dist,0.001)' followed by another casm select --set 'on_hull', one of the positive hull distance configurations is no longer selected. Not sure when the hull_distances are calculated/recalculated, but this is definitely inconsistent behavior.

2. The other structure with non-zero hull distance is just a different sized version of the corresponding hull_structure (slightly different energy cause of numerical things vs the larger supercell + potentially run on different clusters). The casm project knows that they map to each other so both are selected when selecting 'on_hull". Is this expected behavior?

[bpuchala]

I think the seeming inconsistency in (1) is arising because the hull queries have a selection input argument that defaults to MASTER, which is changing in the process of the selections. If you specify all the arguments explicitly using ALL or some fixed selection, then the results should not change.

The hull behavior with non-primitive configurations can be confusing due to the unclear handling of points that are approximately equal. It may be preferable to create a selection of primitive configurations first and do the hull analysis on that.