MathOpt: python_math_opt_solver_test failed on macOS and Windows

[Mizux]

What version of OR-Tools and what language are you using?
Version: main and v99bugfix
OS: MacOS
Language: Python

Which solver are you using (e.g. CP-SAT, Routing Solver, GLOP, BOP, Gurobi)
MathOpt

What did you do?
Build and running python on macOS...

cmake -S. -Bbuild -DBUILD_PYTHON=ON -DBUILD_EXAMPLES=OFF -DBUILD_SAMPLES=OFF
cmake --build build -j 8
(cd build && ctest -C Release --output-on-failure -R "python_math_opt_solver_test")

What did you expect to see
all test passed

What did you see instead?
runner trace:

139/456 Test #139: python_math_opt_solver_test ................................................***Failed    0.63 sec
============================= test session starts ==============================
platform darwin -- Python 3.12.7, pytest-8.3.4, pluggy-1.5.0
rootdir: /Users/runner/work/or-tools/or-tools
collected 15 items

../../../ortools/math_opt/core/python/solver_test.py ..FF...........     [100%]

=================================== FAILURES ===================================
_________ PybindSolverTest.test_invalid_input_throws_error_with_solver _________

self = <solver_test.PybindSolverTest testMethod=test_invalid_input_throws_error_with_solver>
use_solver_class = True

    @parameterized.named_parameters(
        dict(testcase_name="without_solver", use_solver_class=False),
        dict(testcase_name="with_solver", use_solver_class=True),
    )
    def test_invalid_input_throws_error(self, use_solver_class: bool) -> None:
        model = _build_simple_model()
        # Add invalid variable id to cause MathOpt model validation error.
        model.objective.linear_coefficients.ids.append(7)
        model.objective.linear_coefficients.values.append(2.0)
        with self.assertRaisesRegex(StatusNotOk, "id 7 not found"):
>           _solve_model(model, use_solver_class=use_solver_class)

../../../ortools/math_opt/core/python/solver_test.py:115: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

    def _solve_model(
        model: model_pb2.ModelProto,
        *,
        use_solver_class: bool,
        solver_type: parameters_pb2.SolverTypeProto = parameters_pb2.SOLVER_TYPE_GLOP,
        solver_initializer: parameters_pb2.SolverInitializerProto = parameters_pb2.SolverInitializerProto(),
        parameters: parameters_pb2.SolveParametersProto = parameters_pb2.SolveParametersProto(),
        model_parameters: model_parameters_pb2.ModelSolveParametersProto = model_parameters_pb2.ModelSolveParametersProto(),
        message_callback: Optional[Callable[[Sequence[str]], None]] = None,
        callback_registration: callback_pb2.CallbackRegistrationProto = callback_pb2.CallbackRegistrationProto(),
        user_cb: Optional[
            Callable[[callback_pb2.CallbackDataProto], callback_pb2.CallbackResultProto]
        ] = None,
        interrupter: Optional[solver.SolveInterrupter] = None,
    ) -> result_pb2.SolveResultProto:
        """Convenience function for both types of solve with parameter defaults."""
        if use_solver_class:
>           pybind_solver = solver.new(
                solver_type,
                model,
                solver_initializer,
            )
E           RuntimeError: INVALID_ARGUMENT: id 7 not found; Objective.linear_coefficients.ids not found in Variables.ids; ModelProto.objective is invalid

../../../ortools/math_opt/core/python/solver_test.py:59: RuntimeError
_______ PybindSolverTest.test_invalid_input_throws_error_without_solver ________

self = <solver_test.PybindSolverTest testMethod=test_invalid_input_throws_error_without_solver>
use_solver_class = False

    @parameterized.named_parameters(
        dict(testcase_name="without_solver", use_solver_class=False),
        dict(testcase_name="with_solver", use_solver_class=True),
    )
    def test_invalid_input_throws_error(self, use_solver_class: bool) -> None:
        model = _build_simple_model()
        # Add invalid variable id to cause MathOpt model validation error.
        model.objective.linear_coefficients.ids.append(7)
        model.objective.linear_coefficients.values.append(2.0)
        with self.assertRaisesRegex(StatusNotOk, "id 7 not found"):
>           _solve_model(model, use_solver_class=use_solver_class)

../../../ortools/math_opt/core/python/solver_test.py:115: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

    def _solve_model(
        model: model_pb2.ModelProto,
        *,
        use_solver_class: bool,
        solver_type: parameters_pb2.SolverTypeProto = parameters_pb2.SOLVER_TYPE_GLOP,
        solver_initializer: parameters_pb2.SolverInitializerProto = parameters_pb2.SolverInitializerProto(),
        parameters: parameters_pb2.SolveParametersProto = parameters_pb2.SolveParametersProto(),
        model_parameters: model_parameters_pb2.ModelSolveParametersProto = model_parameters_pb2.ModelSolveParametersProto(),
        message_callback: Optional[Callable[[Sequence[str]], None]] = None,
        callback_registration: callback_pb2.CallbackRegistrationProto = callback_pb2.CallbackRegistrationProto(),
        user_cb: Optional[
            Callable[[callback_pb2.CallbackDataProto], callback_pb2.CallbackResultProto]
        ] = None,
        interrupter: Optional[solver.SolveInterrupter] = None,
    ) -> result_pb2.SolveResultProto:
        """Convenience function for both types of solve with parameter defaults."""
        if use_solver_class:
            pybind_solver = solver.new(
                solver_type,
                model,
                solver_initializer,
            )
            return pybind_solver.solve(
                parameters,
                model_parameters,
                message_callback,
                callback_registration,
                user_cb,
                interrupter,
            )
        else:
>           return solver.solve(
                model,
                solver_type,
                solver_initializer,
                parameters,
                model_parameters,
                message_callback,
                callback_registration,
                user_cb,
                interrupter,
            )
E           RuntimeError: INVALID_ARGUMENT: id 7 not found; Objective.linear_coefficients.ids not found in Variables.ids; ModelProto.objective is invalid

../../../ortools/math_opt/core/python/solver_test.py:73: RuntimeError
=========================== short test summary info ============================
FAILED ../../../ortools/math_opt/core/python/solver_test.py::PybindSolverTest::test_invalid_input_throws_error_with_solver - RuntimeError: INVALID_ARGUMENT: id 7 not found; Objective.linear_coefficients.ids not found in Variables.ids; ModelProto.objective is invalid
FAILED ../../../ortools/math_opt/core/python/solver_test.py::PybindSolverTest::test_invalid_input_throws_error_without_solver - RuntimeError: INVALID_ARGUMENT: id 7 not found; Objective.linear_coefficients.ids not found in Variables.ids; ModelProto.objective is invalid
========================= 2 failed, 13 passed in 0.36s =========================

        Start 140: python_math_opt_compute_infeasible_subsystem_result_test

ref: https://github.com/google/or-tools/actions/runs/12116420023/job/33776757913#step:9:318

[lokashrinav]

I might be wrong, but I think this can be fixed by converting the internal error into the custom StatusNotOk Python exception that the test is looking for since its coming through as a generic RuntimeError.

I think we need to add this on line 83 of 'or-tools/ortools/math_opt/core/python/solver_test.py' in the _solve_model function:

except RuntimeError as e:
    raise StatusNotOk(str(e))

[Mizux]

Need to investigate further but the issue seems to come from pybind_abseil directly which didn't wrap the exception the same way on macOS...
I'm currently adding some CI jobs to pybind_abseil project itself and can reproduce the same behaviour so the issue lie in our dependency imho.
see: https://github.com/pybind/pybind11_abseil/actions/runs/12713073661/job/35440135007

[alessio-locatelli]

The issue (closed as duplicate) #4606 happens on Linux, so maybe a more correct issue title could be "on macOS" "on macOS and Linux".