Bug with scipy.linprog feasibility? (A_ub @ x0 <= b_ub).all() is True —but— linprog(np.zeros_like(x0), A_ub=A_ub, b_ub=b_ub) infeasible

With numpy, scipy at versions

numpy 1.25.0
scipy 1.11.0

the following scipy.optimize.linprog call,

import numpy as np
from scipy.optimize import linprog

A_ub = np.array(
      [[-0.15729144,  0.29943807,  0.29311432],
       [-1.32475528, -2.1125364 , -1.55138585],
       [ 1.00861965,  0.53283629, -0.14939833],
       [ 1.07581479,  0.164022  , -1.19889684]])

b_ub = -np.ones(4)

print(linprog(np.zeros(3),
        A_ub=A_ub,
        b_ub=b_ub))

return unfeasible status,

       message: The problem is infeasible. (HiGHS Status 8: model_status is Infeasible; primal_status is At lower/fixed bound)
       success: False
        status: 2
           fun: None
             x: None
           nit: 0

But the problem is actually feasible, since

x0 = np.array([ 229.1748166 , -507.05266751,  512.14005547])
print('x0 is feasible?', (A_ub @ x0 <= b_ub).all())

returns True. Shouldn’t linprog returns a feasible point and a different status code and message in this case?

>Solution :

From the https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.linprog.html:

Note that by default lb = 0 and ub = None.

So x0 = np.array([ 229.1748166 , -507.05266751, 512.14005547]) is not feasible.