Gurobi warning due to improper cleanup order

[staadecker]

from pyoptinterface import gurobi

def make_model():
    env = gurobi.Env(empty=True)
    # env.set_raw_parameter("some_key", "some_value")
    env.start()
    return gurobi.Model(m)

m = make_model()
# m.close() ## Uncommenting this line fixes the issue.

The above code outputs the following warning when the program exits:

Warning: environment still referenced so free is deferred (Continue to use WLS)

This could be avoided by closing the model before the environment is freed. It would be nice for pyoptinterface to do this so that users don't need to.

Note: this only happens with using Gurobi's Web License Server (WLS) not when using a named academic license.

[metab0t]

import gurobipy as gp

options = {
    "WLSACCESSID": "",
    "WLSSECRET": "",
    "LICENSEID": 0,
}

# use wls
# env = gp.Env(params=options)

# use local license
env = gp.Env()

model = gp.Model(env=env)

env.dispose()

The problem is that Gurobi chooses to keep the environment alive if its created model is still alive only for WLS. I do not think we should add more complexity to deal with this issue like recording the models created by which environment and destroy them before the env.

Like the official docs said: "Users should dispose of all models created in this environment before disposing of this Env object."

[staadecker]

Fair enough!