On 05/02/2012 10:51 AM, Fred Mesnard wrote:

Hi there!

I am using PPL 1.0pre1 (GIT version) with SWI-Prolog 6.1.3 and I think I find a bug wrt PIP:

Call: (8) ppl_new_PIP_Problem_from_space_dimension(6, _G10879) ? skip
Exit: (8) ppl_new_PIP_Problem_from_space_dimension(6, 935460) ? creep

Call: (8) ppl_PIP_Problem_add_constraints(935460, [A-B-C+D>=0, F= -A+B+E, D= -1-A+B+C]) ? creep
Exit: (8) ppl_PIP_Problem_add_constraints(935460, []) ?

Call: (8) ppl_PIP_Problem_is_satisfiable(935460) ? skip
Exit: (8) ppl_PIP_Problem_is_satisfiable(935460) ?

IMO, this last call should fail because the [A-B-C+D>=0, F= -A+B+E, D= -1-A+B+C] is not satisfiable. Note that [A-B-C+D>=0, D= -1-A+B+C] is correctly reported as unsatisfiable.

Do you confirm it's a bug from PIP?

Thanks in advance, Fred

Hi Fred.

I confirm it is a bug in PIP_Problem (nothing to do with the Prolog interface itself). The bug was due to lack of update for the `sign' of the special equality row after this row is modified (when elaborating equality constraints after the first one).

I have already committed the correction on master. Can you please have a try on it?

Thanks again for reporting the problem, Enea.