… eaddrinuse errors.
…at's going on.
…e can use it elsewhere.
* New keys_fsm requires ack_keys when backpressure is turned on. * Correct the startup order of applications based on reltool's computed list. * Use the memory backend and its reset() function instead of deleting keys. * Isolate the state directory of the test to .eunit/keys_fsm_eqc. * Cleanup ring files in the state directory before the test instead of rejiggering the ring. * Use a consistent node name and use longnames so that the put_fsm doesn't get confused. * Wait for KV application and service to start rather than using an arbitrary sleep.
The original design for async folding support had the backend make the decision to use async folding at start time, but we changed to having the decision be made per fold call. This change removes some leftover cruft from that original path that is now obsolete.
…e#48) It is possible that the coverage fsm handling the index query at the head of a pipe will finish (or otherwise exit) before we try to link to it. This patch wraps the link in a try-catch to prevent a 'noproc' exit (if the coverage fsm did finish successfully faster than expected, we want to pick up the 'done' message from our inbox). The listkeys pipe coverage fitting had the same potential problem.