BIND 10 #335: b10-xfrout not exiting

[BIND 10 Development]

#335: b10-xfrout not exiting
-------------------------+--------------------------------------------------
      Reporter:  jreed   |        Owner:  vorner   
          Type:  defect  |       Status:  reviewing
      Priority:  major   |    Milestone:           
     Component:  xfrout  |   Resolution:           
      Keywords:          |    Sensitive:  0        
Estimatedhours:  0.0     |        Hours:  0        
      Billable:  1       |   Totalhours:  0        
      Internal:  0       |  
-------------------------+--------------------------------------------------
Changes (by zzchen_pku):

  * owner:  zhanglikun => vorner


Comment:

 Replying to [comment:12 vorner]:
 > The spurious wakeup does not happen usually, as I gather it happens only
 under extreme conditions (high load, bad checksum on incoming packet) and
 it is considered a bug. But as it should not break on working system, it
 is OK to provide a workaround I think. Thanks for including it.
 Got it, thanks.
 > About the race condition, the sleep(1) does help sometime, but not
 always, I was still able to reproduce it. And I found another one. If
 run_timers is run in one thread and shutdown soon after, the shutdown
 calls _is_runnable = False first and then the other thread sets
 _is_runnable = True. May I suggest using a wrapper to run the run_timers
 in another thread, that would store the thread object and calling .join on
 it instead of having the event object? It seems cleaner and would remove
 both problems, because the _is_runnable could be called before the thread
 is started.
 Actually, Zonemgr calls _start_zone_refresh_timer() to run run_timers and
 .join on shutdown, you can find them in Zonemgr class.
 I have updated the test case to call .join on timer thread shutdown. This
 blocks the main thread until the timer is terminated.
 > And I noticed one more thing. We shouldn't call run_timers, then
 shutdown and then run_timers again, because the shutdown pipe would still
 contain b'shutdown' and would still be readable ‒ and we would get a busy
 loop. Maybe it should be mentioned in documentation that one should not
 call run_timers multiple times on the same object. Or, clean the pipe (or
 create a new one and throw the old one away) on each run_timers start.
 Zonemgr process only call run_timers once on startup, but since unittest
 may call it more than once, clean the pipe on each run_timers shutdown to
 handle the busy loop.

 diff :   svn diff -r 3152

-- 
Ticket URL: <http://bind10.isc.org/ticket/335#comment:13>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development