NLopt Fails for Unbounded Contraint Rows

[cohnt]

What happened?

If the bounds for a row of a constraint are vacuous (-∞,∞), then NLopt terminates with kSolverSpecificError. Simple replication below:

import numpy as np
from pydrake.solvers import IpoptSolver, MathematicalProgram, NloptSolver

def f(x):
    return np.array([x[0] + x[1], x[0] - x[1]])

prog = MathematicalProgram()
x = prog.NewContinuousVariables(2)
prog.AddCost(x.dot(x))
prog.AddConstraint(f, lb=[1.0, -np.inf], ub=[1.0, np.inf], vars=x)
prog.SetInitialGuess(x, [0.0, 0.0])

print(NloptSolver().Solve(prog).get_solution_result())
print(IpoptSolver().Solve(prog).get_solution_result())

Version

e45193f

What operating system are you using?

Other, Ubuntu 24.04 (Noble)

What installation option are you using?

compiled from source code using Bazel

Relevant log output

[sherm1]

Assigned to Hongkai for disposition.