GDPopt LBB treats solver RuntimeError as node infeasibility #4023

Activity

1. bernalde commented on Aug 18, 2026

   

   bernalde

   on Aug 18, 2026

   ContributorAuthor

   More actions

   Reproduction confirmed

   Confirmed on fresh Pyomo/pyomo main at 2744a2446cc58598fb9bb53d41ad1b4bc5e86183: both broad RuntimeError catches described in the issue suppress an indeterminate solver failure.

   Environment and command:

   Python 3.12.13
   /home/bernalde/repos/pyomo/.pixi/envs/mindtpy-open/bin/python -m pytest -q -ra pyomo/contrib/gdpopt/tests/test_lbb_solver_errors.py

   Proposed path: pyomo/contrib/gdpopt/tests/test_lbb_solver_errors.py

   # ____________________________________________________________________________________
   #
   # Pyomo: Python Optimization Modeling Objects
   # Copyright (c) 2008-2026 National Technology and Engineering Solutions of Sandia, LLC
   # Under the terms of Contract DE-NA0003525 with National Technology and Engineering
   # Solutions of Sandia, LLC, the U.S. Government retains certain rights in this
   # software.  This software is distributed under the 3-clause BSD License.
   # ____________________________________________________________________________________
   
   """Regression tests for solver-infrastructure failures in GDPopt LBB."""
   
   from unittest.mock import patch
   
   import pyomo.common.unittest as unittest
   import pyomo.contrib.gdpopt.branch_and_bound as branch_and_bound
   from pyomo.contrib.gdpopt.branch_and_bound import GDP_LBB_Solver
   from pyomo.contrib.gdpopt.create_oa_subproblems import (
       add_algebraic_variable_list,
       add_util_block,
   )
   from pyomo.environ import ConcreteModel, Objective, Var, minimize
   from pyomo.opt import SolverResults
   
   
   class _FailingSolver:
       def solve(self, model, **kwds):
           raise RuntimeError("simulated solver infrastructure failure")
   
   
   class TestGDPoptLBBSubproblemErrors(unittest.TestCase):
       def _make_solver_model_and_config(self):
           model = ConcreteModel()
           model.x = Var(bounds=(0, 1))
           model.objective = Objective(expr=model.x)
   
           solver = GDP_LBB_Solver()
           solver.pyomo_results = SolverResults()
           solver.pyomo_results.problem.sense = minimize
           solver.original_util_block = add_util_block(model)
           add_algebraic_variable_list(solver.original_util_block)
   
           config = solver.CONFIG()
           config.minlp_solver = "failing_solver"
           config.minlp_solver_args = {}
           config.local_minlp_solver = "failing_solver"
           config.local_minlp_solver_args = {}
           config.integer_tolerance = 1e-5
           config.time_limit = None
           return solver, model, config
   
       @patch.object(
           branch_and_bound, "SolverFactory", return_value=_FailingSolver()
       )
       def test_evaluated_node_solver_error_propagates(self, solver_factory):
           solver, model, config = self._make_solver_model_and_config()
   
           with self.assertRaisesRegex(
               RuntimeError, "simulated solver infrastructure failure"
           ):
               solver._solve_rnGDP_subproblem(model, config)
   
       @patch.object(
           branch_and_bound, "SolverFactory", return_value=_FailingSolver()
       )
       def test_screening_solver_error_propagates(self, solver_factory):
           solver, model, config = self._make_solver_model_and_config()
   
           with self.assertRaisesRegex(
               RuntimeError, "simulated solver infrastructure failure"
           ):
               solver._solve_local_rnGDP_subproblem(model, config)
   
   
   if __name__ == "__main__":
       unittest.main()

   Red output on unchanged main:

   FF                                                                       [100%]
   E       AssertionError: RuntimeError not raised
   E       AssertionError: RuntimeError not raised
   2 failed in 0.82s

   Path	Observed	Expected
   Evaluated node	Logs “Treating as infeasible” and returns (inf, inf)	Surface an indeterminate solver failure; do not prune the node as mathematically infeasible
   Local screening	Logs “Treating as infeasible” and returns (-inf, inf)	Surface the failed screening solve instead of silently continuing

   Non-vacuity check: adding a throwaway raise at the start of each catch made this exact test file pass (2 passed in 0.21s); the source change was then discarded.

   Acceptance criteria

   • A node-solver exception that does not prove mathematical infeasibility cannot produce infeasible node bounds or prune a branch.
   • The evaluated-node and optional local-screening paths both surface solver/infrastructure failures through an explicit failure contract.
   • Existing handling for FBBT-proven infeasibility and TerminationCondition.infeasible remains intact.
   • The regression above and the existing GDPopt LBB tests pass.