There are several problems with the approach to testing; admittedly, Jabber isn't an easy protocol to test, but I believe every single one of the features involved in xmpp4r-simple should be mocked to include no networking, and certainly no real Jabber accounts.
This would resolve intermittent errors due to network latency et al. It would also make writing further tests much simpler (and therefore more probable). Also, it will help with having to deal with side-effects of previous tests (if there are any) due to the external dependency.
Personally, it's painful and disheartening to sit there with the tests failing due to network problems, especially when they sit there for 20 seconds for each network-sensitive test.
There are several problems with the approach to testing; admittedly, Jabber isn't an easy protocol to test, but I believe every single one of the features involved in xmpp4r-simple should be mocked to include no networking, and certainly no real Jabber accounts.
This would resolve intermittent errors due to network latency et al. It would also make writing further tests much simpler (and therefore more probable). Also, it will help with having to deal with side-effects of previous tests (if there are any) due to the external dependency.
Personally, it's painful and disheartening to sit there with the tests failing due to network problems, especially when they sit there for 20 seconds for each network-sensitive test.