Strange behavior of acyclic/2 constraint

[Peter Bernschneider]

When solving the Janko Wamuzu puzzle I had used a matrix M containing arrows to describe the paths between respectively two circle squares. For instance 1 I got a result containing cycles. I then included the acyclic constraint (see line 29) and found the correct solution (see wamuzu_en.pi).

But for instance 17 the SAT solver found no solution. I then removed the acyclic constraint and got the correct solution, which in my eyes is acyclic anyway (see attached screen shot).

Is it possible, that acyclic/2 has a bug?

Cheers, Peter

[Neng-Fa Zhou]

Hi Peter,

Thank you for the program. Yes, your program reveals a bug in the newly added constraint acyclic_d. The decomposer of the constraint over-constrains when there are side constraints involved in addition to acyclicity. The bug has been fixed and a patched version (3.6#5) has been released.

Cheers,

NF