}
        },
        directExecutor());
    manager.startAsync();
    afterStarted.countDown();
    // We do not call awaitHealthy because, due to races, that method may throw an exception.  But
    // we really just want to wait for the thread to be in the failure callback so we wait for that
    // explicitly instead.
    failEnter.await();

      new AtomicReference<>(immediateVoidFuture());

  private @LazyInit ThreadConfinedTaskQueue latestTaskQueue = new ThreadConfinedTaskQueue();

  /**
   * This object is unsafely published, but avoids problematic races by relying exclusively on the
   * identity equality of its Thread field so that the task field is only accessed by a single
   * thread.
   */
  private static final class ThreadConfinedTaskQueue {
    /**

    the certificate that follows and the last certificate is signed by a trusted root.

 *  Fix: Don't lose HTTP/2 flow control bytes when incoming data races with a stream close. If this
    happened enough then eventually the connection would stall.

 *  Fix: Acknowledge and apply inbound HTTP/2 settings atomically. Previously we had a race where we

    `Cache.urls()` would prevent in-flight entries from being written.


## Version 3.14.3

_2019-09-10_

 *  Fix: Don't lose HTTP/2 flow control bytes when incoming data races with a stream close. If this
    happened enough then eventually the connection would stall.

 *  Fix: Acknowledge and apply inbound HTTP/2 settings atomically. Previously we had a race where we

    assertEquals(getKey + suffix, map.get(getKey));
    assertEquals(refreshKey + suffix, map.get(refreshKey));
  }

  @SuppressWarnings("ThreadPriorityCheck") // doing our best to test for races
  public void testExpandDuringLoading() throws InterruptedException {
    final int count = 3;
    final AtomicInteger callCount = new AtomicInteger();
    // tells the computing thread when to start computing

							SkipDecommissioned: true, // make sure we skip the decommissioned pool
							NoAuditLog:         true,
						})
					if err != nil {
						// This can happen when rebalance stop races with ongoing rebalance workers.
						// These rebalance failures can be ignored.
						if isErrObjectNotFound(err) || isErrVersionNotFound(err) || isDataMovementOverWriteErr(err) {
							ignore = true
							stopFn(0, nil)

   * listeners for future n may complete before some for future n-1.) However, it is possible, if
   * one input completes with result X and another later with result Y, for Y to come before X in
   * the output future list. (Such races are impossible to solve without global synchronization of
   * all future completions. And they should have little practical impact.)
   *

   * listeners for future n may complete before some for future n-1.) However, it is possible, if
   * one input completes with result X and another later with result Y, for Y to come before X in
   * the output future list. (Such races are impossible to solve without global synchronization of
   * all future completions. And they should have little practical impact.)
   *

- Client-go: fixed potential data races retrying requests using a custom `io.Reader` body; with this fix, only requests with no body or with `string` / `[]byte` / `runtime.Object` bodies can be retried ([#113933](https://github.com/kubernetes/kubernetes/pull/113933), [@liggitt](https://...

    * The vSphere cloud provider has a cleanup function with a loop,
    * inside of which a write lock was intended to prevent races against
    * VM creation, but the usage of the sync.RWMutex type is incorrect,
    * plus is placed in a defer call so the locking is effectively
    * nonexistent.  This patch adds an anonymous function call inside