Timeouts in pylibmc 1.2.3

[lcosmin]

I'm using the following code, combined with dropping all packets going to port 21111 ( iptables -A INPUT -p tcp --dport 21111 -j DROP):

import pylibmc

c = pylibmc.Client(["localhost:21111"], binary=True, behaviors={ 
        "no_block" : True, 
        "connect_timeout" : 2, 
        "send_timeout" : 5, 
        "receive_timeout" : 5 })

c.add("foo", "bar")

The connect_timeout parameter doesn't seem to work, because the test program hangs for over one minute and finishes with _pylibmc.MemcachedError: error 26 from memcached_add: Operation now in progress .
Is this a known issue or my test case is somehow wrong? I assume the timeouts are in seconds, since the documentation doesn't mention this.

[lericson]

Odd behavior, #56 mentions _poll_timeout, does that help?

[lcosmin]

No, not really... I've tried different combinations of connect_timeout and _poll_timeout , even one without the other, etc. but nothing seems to reduce the timeout to something under 1 minute. binary=False didn't help either.

[dkuebric]

I did some digging on this issue (affecting a project I'm working on) and I'm pretty sure it's a problem in libmemcached. Running against 0.44 (the ubuntu 12.04 default), I had the ignored-connect-timeout problem. However, against a fresh build of libmemcached 1.0-15 it works as intended (or, well, it raises that exception after the expected time).