for prod and one for tests. However, unlike our other packages, as of this
    writing it has test supersource but no prod supersource.

    GWT happens to use the prod .gwt.xml, so it looks for no supersource for
    tests, either. This causes it to fail to find AtomicLongMapTest.

    Our workaround is to tell GWT that util.concurrent and all other packages

  // Use fewer steps in the ListIteratorTester in ListListIteratorTester because it's slow in prod
  // mode.
  static int listListIteratorTesterNumIterations() {
    // TODO(hhchan): It's 4 in java.  Figure out why even 3 is too slow in prod mode.
    return 2;
  }

  // Use fewer steps in the IteratorTester in CollectionIteratorTester because it's slow in prod
  // mode.
  static int collectionIteratorTesterNumIterations() {
    return 3;
  }

 * doing so does <i>not</i> give a reliable measurement of elapsed time, because wall time readings
 * are inherently approximate, routinely affected by periodic clock corrections. Because this class
 * (by default) uses {@link System#nanoTime}, it is unaffected by these changes.
 *
 * <p>Use this class instead of direct calls to {@link System#nanoTime} for two reasons:
 *
 * <ul>

     * <p>{@link #transitioned}: is set by {@link #transitionService} to indicate that some
     * transition has been performed.
     *
     * <p>Together, they allow us to enforce that all services have their listeners installed prior
     * to any service performing a transition, then we can fail in the ServiceManager constructor
     * rather than in a Service.Listener callback.
     */
    @GuardedBy("monitor")
    boolean ready;

    executor.execute(task);
    assertEquals(1, numCalls.get());
  }

  /*
   * Under Android, MyError propagates up and fails the test?
   *
   * TODO(b/218700094): Does this matter to prod users, or is it just a feature of our testing
   * environment? If the latter, maybe write a custom Executor that avoids failing the test when it
   * sees an Error?
   */
  @AndroidIncompatible

 * requirement that existing callers already fulfill).
 *
 * Disclaimer: Part of the reason that callers are so well adapted to `Function<A, B>` may be that
 * that is how the signature looked even prior to this comment! So naturally any change can break
 * existing users, but it can't *fix* existing users because any users who needed
 * `Function<@Nullable A, @Nullable B>` already had to find a workaround. Still, there is a *ton* of

      if (result == CancellationException.class) {
        assertTrue(future.isCancelled());
        assertTrue(cancellationSuccess.get());
        // cancellation can interleave in 3 ways
        // 1. prior to setFuture
        // 2. after setFuture before set() on the future assigned
        // 3. after setFuture and set() are called but before the listener completes.

child was very fond of pretending to be two people.  `But it's no
use now,' thought poor Alice, `to pretend to be two people!  Why,
there's hardly enough of me left to make ONE respectable
person!'

  Soon her eye fell on a little glass box that was lying under
the table:  she opened it, and found in it a very small cake, on
which the words `EAT ME' were beautifully marked in currants.

     */
    void postReadCleanup() {
      if ((readCount.incrementAndGet() & DRAIN_THRESHOLD) == 0) {
        cleanUp();
      }
    }

    /**
     * Performs routine cleanup prior to executing a write. This should be called every time a write
     * thread acquires the segment lock, immediately after acquiring the lock.
     *
     * <p>Post-condition: expireEntries has been run.
     */

      // threshold is equal to ceilingPowerOfTwo(expectedSize). There is a separate code
      // path when the first operation on the new map is putAll(otherMap). There, prior to
      // https://github.com/openjdk/jdk/commit/3e393047e12147a81e2899784b943923fc34da8e, a bug
      // meant that sometimes a too-large threshold is calculated. However, this new threshold is