psycogreen breaks connect_timeout option

[redlolgeerf]

python version 3.8.6

library versions:
gevent==20.9.0
greenlet==0.4.17
psycogreen==1.0.2
psycopg2==2.8.6

minimal example:

import gevent
import gevent.monkey

gevent.monkey.patch_all()  # noqa

import psycogreen.gevent

psycogreen.gevent.patch_psycopg()  # noqa

import psycopg2


def fetch(num, secs):
    print("connecting to db")
    conn = psycopg2.connect("dbname=postgres connect_timeout=1")
    print("connected")


jobs = [gevent.spawn(fetch, 3, 2)]
gevent.joinall(jobs)

This will print "connecting to db" and hang. If you comment out line psycogreen.gevent.patch_psycopg() # noqa code will fail with OperationalError in couple seconds.

[redlolgeerf]

Desired behavior can be achieved, if I replace psycogreen.gevent.patch_psycopg() with

from psycogreen.gevent import gevent_wait_callback
from psycopg2 import extensions
from functools import partial
extensions.set_wait_callback(partial(gevent_wait_callback, timeout=1))

Maybe, this should be added to documentation.

[dvarrazzo]

Yes, I'm aware: in async mode the libpq doesn't respect the connect_timeout, it is the client which should re-implement it. So this is actually a psycopg2 problem. There might be a ticket open for it already... yup, this: psycopg/psycopg2#944

[redlolgeerf]

Thanks, I'll close this issue then.