[tor-bugs] #5917 [Stem]: integ.control.base_controller hangs

[Tor Bug Tracker & Wiki]

#5917: integ.control.base_controller hangs
---------------------+------------------------------------------------------
 Reporter:  gsathya  |          Owner:  atagar
     Type:  defect   |         Status:  new   
 Priority:  normal   |      Milestone:        
Component:  Stem     |        Version:        
 Keywords:           |         Parent:        
   Points:           |   Actualpoints:        
---------------------+------------------------------------------------------

Comment(by atagar):

 I've reproed this on a mac and spent the last four hours unsuccessfully
 trying to figure it out. Here's what I know so far...

 - The issue always occurs in test_status_notifications(), when we're
 sending garbage data to the socket prior to calling AUTHENTICATE (which
 causes tor to terminate the socket connection). We're hanging up on the
 readline() call when trying to receive the response, which should *not* be
 blocking (it should either read content or error).

 - This issue only manifests if we also run a couple other base_controller
 integ tests that put a heavy load on the socket prior to the problem. From
 what I can tell nothing should be bleeding between the tests, so my guess
 is that the prior activity must be putting the socket into some
 problematic state.

 - This occurs with both port and file based sockets.

 I'm using a socket based file object (derived from socket.makefile()) to
 have an efficient method for reading newline delimited responses. I
 haven't a clue why it's only blocking on macs, and only after encountering
 a heavy load (python bug?).

 Guess that tomorrow I'll try removing our usage of a socket file, and make
 raw recv() calls to see if that works.

-- 
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/5917#comment:12>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online