In edu.umd.cs.guitar.testcase.plugin.TCPlugin:
parseFollowRelations(): I couldn't locate the explicit semantics for a
'reaching' vs. a 'follow' edge, but it looks like the reaching edge comes
from a hidden widget to a non-hidden widget. However, the back-trace in
DFS() uses the reaching edges *only* (based upon how parseFollowRelations()
populates the 'pred' map data structure), which leads to a number of
problems, especially when SequenceLengthCoverage is running (it seems that
a number of paths are not found that could lead to root).