Model becomes infeasible in version 9.15 #5003
Description
What version of OR-Tools and what language are you using?
Version:
- OR-Tools 9.14 (baseline; model solves and is feasible)
- OR-Tools 9.15 (regression; triggers “Infeasible solution!” / failure)
Language: Python
Which solver are you using (e.g. CP-SAT, Routing Solver, GLOP, BOP, Gurobi)
CP-SAT
What operating system (Linux, Windows, ...) and version?
Windows 11
What did you do?
Steps to reproduce the behavior:
- Create a Python environment with OR-Tools 9.15.
- Load the attached CP-SAT model (attached as pbtxt:
my_model.pbtxt). - Solve it with CP-SAT using the following key parameter settings:
max_time_in_seconds = 2num_search_workers = 1
- Observe that the solver reports an “Infeasible solution!” and dumps a response to
wrong_response.pb.txt.
(For comparison) Repeat the same procedure using OR-Tools 9.14, and observe the model solves normally and is feasible.
What did you expect to see
Consistent behavior across versions: the model should be feasible in OR-Tools 9.15, as it is in 9.14 (and in my production use prior to upgrading).
What did you see instead?
In OR-Tools 9.15, solving fails with an “Infeasible solution!” message and a stack trace:
Solution 18, time = 0.76 s, objective = 3122.0
Infeasible solution! source: 'main' dumped CpSolverResponse to '.\wrong_response.pb.txt'.
*** Check failure stack trace: ***
@ 00007FFD4DEC48AC (unknown)
@ 00007FFD4E162C66 (unknown)
@ 00007FFD4DED5554 (unknown)
@ 00007FFD4E15473B (unknown)
@ 00007FFD6467A487 (unknown)
@ 00007FFD647210A8 (unknown)
@ 00007FFD647211D4 (unknown)
@ 00007FFD6465D21C (unknown)
@ 00007FFD69D1F5F8 (unknown)
@ 00007FFD69D66B93 (unknown)
@ 00007FFD69D664F5 (unknown)
@ 00007FFD69D6BDE6 (unknown)
@ 00007FFD69D130F4 (unknown)
@ 00007FFD69D08ED5 (unknown)
@ 00007FFD69D091D5 (unknown)
@ 00007FFD69D092E0 (unknown)
@ 00007FFD69CE1E57 (unknown)
@ 00007FFD69CE19A4 (unknown)
@ 00007FFD69CE29A5 (unknown)
@ 00007FFD69CE2F36 (unknown)
@ 00007FFD69CE2FC3 (unknown)
@ 00007FFD69CE4BF1 (unknown)
@ 00007FFD69CE49C1 (unknown)
@ 00007FFD69CE52E9 (unknown)
@ 00007FF712B61230 (unknown)
@ 00007FFDF212E8D7 (unknown)
@ 00007FFDF2ECC53C (unknown)
Additionally: the dumped response file (wrong_response.pb.txt) appears to indicate FEASIBLE in OR-Tools 9.15, but the run still aborts with “Infeasible solution!”. I do not understand why the solver indicates feasibility yet triggers an infeasible-solution failure.