#8347 BLOC 8.2.0 (: 5 instances of Clock Activity eventually all die (from oom-killer ?)

[Zarro Boogs per Child]

#8347: 5 instances of Clock Activity eventually all die (from oom-killer ?)
-----------------------------------+----------------------------------------
 Reporter:  thomaswamm             |         Owner:  dsd                              
     Type:  defect                 |        Status:  new                              
 Priority:  blocker                |     Milestone:  8.2.0 (was Update.2)             
Component:  stability-reliability  |       Version:  Development build as of this date
 Keywords:                         |   Next_action:  reproduce                        
 Verified:  0                      |     Blockedby:                                   
 Blocking:                         |  
-----------------------------------+----------------------------------------
 Testing beta build 8.2-759, trying to find simpler test cases to prove or
 diagnose system instability.

 I restart my G1G1 XO, then launch 5 instances of Clock-2 activity, each
 with a different appearance (right-click 0, 1, 2, 3, or 4 times) just so I
 can tell them apart. The XO gets noticeably sluggish which is why I
 stopped at 5 instances.  Then I go away.  A few hours later I come back
 and see all 5 Clocks are gone.

 I tried this with no power management options checkmarked, then again with
 Extreme Power Management checkmarked (i.e. no wlan). Similar final result
 (no clocks), but slightly different logs.

 I think something goes wrong under the hood, oom-killer gets invoked, and
 a cascade ultimately crashes X-server, then X and Sugar are automagically
 restarted.  I am only pretending to know what I am talking about, but I
 have seen MANY similar crashes with other activities, with oom-killer
 showing in /var/log/messages.  So I don't believe this is a bug in Clock
 activity, it is something lower down that impacts all activities.

 In /var/log/messages, the date stamp suggests the crash cascade happens
 within an hour of me launching the 5 Clocks.

 Does anyone have a simpler test case to trigger the bug?  Is more work
 needed to diagnose this problem?  What testing tools are available?  All I
 have is one G1G1 XO, and some time.

-- 
Ticket URL: <http://dev.laptop.org/ticket/8347>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system