import numpy as np
from sklearn.linear_model import LassoLarsIC
def check_lassoIC_positive_trivial(add_noise=True):
        est = LassoLarsIC(fit_intercept=False, positive=True)
        y = np.array([0., 1, 2, 3, 4, 5, 6, 7])
        x = y.copy()
        X = np.c_[x, x*x]
        if add_noise:
            y += np.random.uniform(len(y)) * 0.01
        est.fit(X.copy(), y)
        preds = est.predict(X)
        print(est.coef_)
        return est.coef_



coef = check_lassoIC_positive_trivial(add_noise=True)
assert np.allclose(coef[1], 0.)

# This one fails
coef = check_lassoIC_positive_trivial(add_noise=False)
assert np.allclose(coef[1], 0.)