ZOOKEEPER-XXXX: Restore interrupt status in quorum catch blocks

[TimurRakhmatullin86]

Summary

Multiple catch (InterruptedException) blocks in the quorum package swallow the interrupt flag without calling Thread.currentThread().interrupt(). This violates the Java interrupt contract: any code higher in the call stack that checks Thread.interrupted() or calls a blocking method will never see the interruption, which can cause threads to hang during shutdown or fail to terminate promptly.

Changes

This patch adds Thread.currentThread().interrupt() as the first statement in 9 catch blocks across 6 files:

File	Method / Context
QuorumCnxManager.java	halt() — after listener.join()
QuorumCnxManager.java	ListenerHandler.acceptConnections() — after Thread.sleep() in retry loop
QuorumCnxManager.java	SendWorker.run() — after pollSendQueue()
QuorumPeerMain.java	runFromConfig() — after quorumPeer.join()
LearnerHandler.java	shutdown() — after queuedPackets.put()
Observer.java	waitForReconnectDelayHelper() — after Thread.sleep()
Learner.java	connectToLeader() — after latch.await()
Learner.java	connectToLeader() finally block — after awaitTermination()
FastLeaderElection.java	WorkerReceiver.run() — after manager.pollRecvQueue()

Motivation

Per Java best practices (see Java Concurrency in Practice §7.1.3 and the InterruptedException Javadoc), when a method catches InterruptedException and does not re-throw it, it must restore the interrupt status by calling Thread.currentThread().interrupt(). Without this, the interrupt signal is permanently lost. In ZooKeeper's quorum code, this means:

• Shutdown delays: QuorumCnxManager.halt() and SendWorker.run() may not terminate promptly because callers polling Thread.interrupted() never see the flag.
• Silent failures in leader election: FastLeaderElection.WorkerReceiver and Learner.connectToLeader() continue looping after an interrupt without propagating the signal, potentially delaying or preventing clean shutdown.
• Stale retry loops: ListenerHandler.acceptConnections() catches the interrupt during a sleep-based retry but never restores the flag, so the retry loop does not respect the shutdown signal from higher-level code.

Testing

• No behavioral change: each catch block still logs the same message at the same level.
• The only addition is a single Thread.currentThread().interrupt() call so the flag is preserved for upstream callers.
• Existing test suites pass without modification.

🤖 Generated with Claude Code