Divergent Score Reporting for Solver run

[terbrack]

Hi,
I am facing the following issue regarding the score reporting of a solving run:
there is a mismatch between the solver score printed after solving ended and the constraint list with associated scores, see:

2024-05-07 18:03:14,008 [main] INFO Solving started: time spent (44), best score (0hard/-73572medium/-73226soft), environment mode (REPRODUCIBLE), move thread count (NONE), random (JDK with seed 0).

2024-05-07 18:03:14,016 [main] INFO Construction Heuristic phase (0) ended: time spent (53), best score (0hard/-73572medium/-73226soft), score calculation speed (0/sec), step total (0).

2024-05-07 18:05:15,505 [main] INFO Local Search phase (1) ended: time spent (121542), best score (0hard/-72401medium/-71917soft), score calculation speed (288/sec), step total (13).

2024-05-07 18:05:15,509 [main] INFO Solving ended: time spent (121543), best score (0hard/-72401medium/-71917soft), score calculation speed (288/sec), phase total (2), environment mode (REPRODUCIBLE), move thread count (NONE).

constraint matchCount score
1 constraint_one 3 0hard/0medium/-3soft
2 constraint_two 2 0hard/0medium/-2soft
3 constraint_three 16 0hard/-2811medium/0soft
4 constraint_four 90 0hard/-71831medium/0soft
5 constraint_five 90 0hard/0medium/-71912soft
Explanation of score (0hard/-74642medium/-71917soft)

What makes me wonder is the mismatch between the reported score at the end of local search phase, (0hard/-72401medium/-71917soft) , and the summed up score in the constraint match summary, (0hard/-74642medium/-71917soft).

Do you have an idea why this could happen?
Thanks,
Hajo

[triceo]

Hey there, my first bet is a score corruption. Run for a long time with FULL_ASSERT environment mode and if it fails, you have a bug in your constraints and/or variable listeners and/or custom moves.