Description
When analyzing output of the MasterAllocatorTest.SlaveLost test I spotted following logs:
I1020 18:18:09.026553 237658112 status_update_manager.cpp:322] Received status update TASK_RUNNING (UUID: 767597b2-f9de-464b-ac20-985452a897e6) for task 0 of framework 7aff439d-307c-486b-9c0d-c2a47ddbda5b-0000 I1020 18:18:09.026845 234438656 slave.cpp:3090] Forwarding the update TASK_RUNNING (UUID: 767597b2-f9de-464b-ac20-985452a897e6) for task 0 of framework 7aff439d-307c-486b-9c0d-c2a47ddbda5b-0000 to master@172.18.6.110:62507 I1020 18:18:09.026973 234438656 slave.cpp:651] ; unregistering and shutting down I1020 18:18:09.027007 234438656 slave.cpp:2016] Asked to shut down framework 7aff439d-307c-486b-9c0d-c2a47ddbda5b-0000 by @0.0.0.0:0 I1020 18:18:09.027019 234438656 slave.cpp:2041] Shutting down framework 7aff439d-307c-486b-9c0d-c2a47ddbda5b-0000
It looks like Slave::shutdown() uses wrong assumptions about possible execution paths.