[tor-bugs] #2077 [Tor Client]: Tor crashes in Vidalia 0.2.2.17a-0.2.10

[Tor Bug Tracker & Wiki]

#2077: Tor crashes in Vidalia 0.2.2.17a-0.2.10
--------------------------------+-------------------------------------------
 Reporter:  joebt               |        Type:  defect    
   Status:  needs_review        |    Priority:  normal    
Milestone:  Tor: 0.2.2.x-final  |   Component:  Tor Client
  Version:                      |    Keywords:  Tor,      
   Parent:                      |  
--------------------------------+-------------------------------------------

Comment(by arma):

 The problem here appears to be in the line
 {{{
     global_state->LastWritten = -1;
 }}}
 added by git 29e0d708. special's backtrace is
 {{{
 or_state_save -> config_dump -> get_assigned_option
 (key="LastWritten") -> format_iso_time -> tor_gmtime_r -> crash
 }}}

 My first thought was "Mike meant to set LastWritten to 0".

 My second thought was "wait, why are we using LastWritten to remember
 whether the last write had failed?"

 My third thought was "The behavior of LastWritten in maint-0.2.1 is to put
 time(NULL) into the state file, so the state file tells you when it was
 written. The behavior of LastWritten in maint-0.2.2 is now to tell you
 when the *previous* state file was written. We should reconcile these. I
 rather like the behavior in maint-0.2.1."

 If Mike wants to turn off circuit building when we're failing to write to
 our state file, it seems to me that we should have some wrapper functions
 like last_state_write_worked() and note_state_file_written() that he can
 call.

-- 
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/2077#comment:13>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online