Ring closure bonds missing

[jthorton]

When using Chemper to generate SMARTS patterns which involve rings we have noticed that ring closure bonds are missing.
For example, we are using chemper to make bespoke torsion patterns which include all layers to ensure that the pattern is specific to only one molecule, in the case of OC1CCCC1 chemper produces the following smirks for one of the torsions
[#1AH0X1x0!r+0:1]-!@[#8AH1X2x0!r+0:2]-!@[#6AH1X4x2r5+0:3](-@[#6AH2X4x2r5+0](-!@[#1AH0X1x0!r+0])(-!@[#1AH0X1x0!r+0])-@[#6AH2X4x2r5+0](-!@[#1AH0X1x0!r+0])(-!@[#1AH0X1x0!r+0])-@[#6AH2X4x2r5+0](-!@[#1AH0X1x0!r+0])(-!@[#1AH0X1x0!r+0])-@[#6AH2X4x2r5+0](-!@[#1AH0X1x0!r+0])-!@[#1AH0X1x0!r+0])-!@[#1AH0X1x0!r+0:4] which has no ring clousre bonds, this results in the following structure:

rather than:

This has also been an issue in the openforcefield toolkit see here