[tor-bugs] #3035 [Tor Client]: fix a few clang-static analyzer complaints in 0.2.2

[Tor Bug Tracker & Wiki]

#3035: fix a few clang-static analyzer complaints in 0.2.2
------------------------+---------------------------------------------------
 Reporter:  Sebastian   |          Owner:                    
     Type:  defect      |         Status:  needs_review      
 Priority:  normal      |      Milestone:  Tor: 0.2.2.x-final
Component:  Tor Client  |        Version:                    
 Keywords:              |         Parent:                    
   Points:              |   Actualpoints:                    
------------------------+---------------------------------------------------

Comment(by Sebastian):

 Replying to [comment:10 arma]:
 > Replying to [comment:8 Sebastian]:
 > > > What if tor_assert triggers, causing us to do a log to the
 controller at the same time, which would want to disable logging, ...
 > >
 > > I don't see the issue here, maybe I am silly? We haven't yet stopped
 logging to the controller, so the assert will get logged (we don't disable
 logging to the controller until after the assert, and tor_assert doesn't
 disable control logging) and die.
 >
 > Yeah. I dunno. Maybe? I know tor_assert logs when it triggers, and I
 think it also tries to flush to the controller.

 Yes, it does log but doesn't call CONN_LOG_PROTECT() again, so an infinite
 loop isn't possible here.

 > I'm not smart enough right now to be confident that this is safe. And if
 we currently think the tor_assert is never triggered, but we don't know
 what would happen if it is, we're basically putting down a landmine for
 our future selves.

 I'm not saying our behavior is safe because we don't think we're currently
 not triggering the assert

-- 
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/3035#comment:11>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online