[tor-bugs] #1666 [Tor Client]: SOCKS handling should accept (and ignore) password auth.

[Tor Bug Tracker & Wiki]

#1666: SOCKS handling should accept (and ignore) password auth.
-------------------------+--------------------------------------------------
 Reporter:  nickm        |          Owner:  mwenge            
     Type:  enhancement  |         Status:  needs_review      
 Priority:  normal       |      Milestone:  Tor: 0.2.3.x-final
Component:  Tor Client   |        Version:                    
 Keywords:               |         Parent:                    
   Points:               |   Actualpoints:                    
-------------------------+--------------------------------------------------

Comment(by nickm):

 Let's get this merged!

 For changes please see my new branch "bug1666" in my public repository,
 based on Robert's branch above.

 I think that the want_length_out computations in parse_socks() are
 suspicious.  I've fixed these (I think) in 1ed615ded7db0765e.

 The big list of tests makes me quite happy.  I refactored them into a test
 suite of their own.  Previously, some of the subtests were only passing
 because of state that was already set in the socks_request_t; I fixed
 that.

 I think that the current code is incorrect in how it getting a message
 that starts with  0x01.  Right now, if we have negotiated socks5, then it
 accepts username/password auth.  But it does this potentially infinitely
 many times, and it also does it if we have not negotiated
 username/password auth!  I think that we should be more strict here.  My
 aec396d might fix that.

 In 2e6604f4 I added code to record the username/password; we'll need that.

 In 05c424 I changed the calling convention for fetch_from_buf_socks to
 more closely match fetch_from_evbuffer socks.

 This probably needs another review now, and maybe a little more testing.

-- 
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/1666#comment:13>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online