Closed scijones closed 1 month ago
I went ahead and manually ran the 90 games. It doesn't seem to hang within a game and runs for about the same number of decision cycles as before and ends in the same state -- halted and waiting using the operator "wait-for-response".
So, I think this is fine to merge.
I will also add this as a feature request if and when this is pulled. The problematic tests worth a future second glance:
Teach_Soar_90_Games Demo_ToH_Recursive (probably worth just omitting this test from usual unit testing or rewriting the agent for this test) Operator_Selection_Knowledge (AKA Operator_Selection_Knowledge_Mega_Test) Constraint_Prop_from_Base_Conds
Everything else mostly seemed to be about removing redundant testing in the expected chunks where there really was only one thing in a superstate, but two branches of logic had varying levels of constraint. In these cases, because of the extra constraints now being condensed to constrain one symbol, chunks are essentially the same, but less general hopefully in a more correct way for typical usage.