When killing a node, it may cause a cascade drop for the tasks of this node, where there will be a reentrancy of a lock in the call stack. This way, the program will fall into a deadlock.
This PR fixes it by introducing a temporary store for runnable when being wakened up on kill so that the cascade drop can be broken.
When killing a node, it may cause a cascade drop for the tasks of this node, where there will be a reentrancy of a lock in the call stack. This way, the program will fall into a deadlock. This PR fixes it by introducing a temporary store for
runnable
when being wakened up on kill so that the cascade drop can be broken.