private final transient int size;

  /*
   * We have some considerable complexity in these create methods because of
   * Builder.buildKeepingLast(). The same Builder might be called with buildKeepingLast() and then
   * buildOrThrow(), or vice versa. So in particular, if we modify alternatingKeysAndValues to
   * eliminate duplicate keys (for buildKeepingLast()) then we have to ensure that a later call to

            (workerRunningState == IDLE || workerRunningState == QUEUING)
                && queue.removeLastOccurrence(submittedTask);
        // If the delegate is directExecutor(), the submitted runnable could have thrown a REE. But
        // that's handled by the log check that catches RuntimeExceptions in the queue worker.
        if (!(t instanceof RejectedExecutionException) || removed) {
          throw t;
        }
      }

    /*
     * Since we use AtomicReferenceFieldUpdaterAtomicHelper by default, we'll "obviously" use it
     * even when Unsafe isn't available. But it's nice to have a check here to make sure that
     * nothing somehow goes wrong as the JDK restricts access to Unsafe.
     */
    checkHelperVersion(NO_UNSAFE, "AtomicReferenceFieldUpdaterAtomicHelper");
    /*

  // Timed Get
  // There are a few design constraints to consider
  // * We want to be responsive to small timeouts, unpark() has non trivial latency overheads (I
  //   have observed 12 micros on 64-bit linux systems to wake up a parked thread). So if the
  //   timeout is small we shouldn't park(). This needs to be traded off with the cpu overhead of

  // Timed Get
  // There are a few design constraints to consider
  // * We want to be responsive to small timeouts, unpark() has non trivial latency overheads (I
  //   have observed 12 micros on 64-bit linux systems to wake up a parked thread). So if the
  //   timeout is small we shouldn't park(). This needs to be traded off with the cpu overhead of

   * `INADEQUATE_SECURITY`.
   */
  HTTP_2("h2"),

  /**
   * Cleartext HTTP/2 with no "upgrade" round trip. This option requires the client to have prior
   * knowledge that the server supports cleartext HTTP/2.
   *
   * See also [Starting HTTP/2 with Prior Knowledge][rfc_7540_34].
   *
   * [rfc_7540_34]: https://datatracker.ietf.org/doc/html/rfc7540#autoid-10
   */

  @SuppressWarnings("Immutable")
  private final long[] array;

  /*
   * TODO(kevinb): evaluate the trade-offs of going bimorphic to save these two fields from most
   * instances. Note that the instances that would get smaller are the right set to care about
   * optimizing, because the rest have the option of calling `trimmed`.
   */

  private final transient int start; // it happens that we only serialize instances where this is 0

  @SuppressWarnings("Immutable")
  private final int[] array;

  /*
   * TODO(kevinb): evaluate the trade-offs of going bimorphic to save these two fields from most
   * instances. Note that the instances that would get smaller are the right set to care about
   * optimizing, because the rest have the option of calling `trimmed`.
   */

  private final transient int start; // it happens that we only serialize instances where this is 0

   * filters to be compatible, they must:
   *
   * <ul>
   *   <li>not be the same instance
   *   <li>have the same number of hash functions
   *   <li>have the same bit size
   *   <li>have the same strategy
   *   <li>have equal funnels
   * </ul>
   *
   * @param that The Bloom filter to check for compatibility.
   * @since 15.0
   */

                   (which is good: our tests have intentional violations), but
                   Error Prone doesn't know it's building test code unless we
                   pass -XepCompilingTestOnlyCode, and that argument needs to
                   be passed as part of the same <arg> as -Xplugin:ErrorProne,
                   and I gave up trying to figure out how to do that for test