*     value.)
   */
  @CanIgnoreReturnValue
  @CheckForNull
  V forcePut(@ParametricNullness K key, @ParametricNullness V value);

  // Bulk Operations

  /**
   * {@inheritDoc}
   *
   * <p><b>Warning:</b> the results of calling this method may vary depending on the iteration order
   * of {@code map}.
   *
   * @throws IllegalArgumentException if an attempt to {@code put} any entry fails. Note that some

   * That is, each node's subtrees are traversed after the node itself is returned.
   *
   * <p>No guarantees are made about the behavior of the traversal when nodes change while iteration
   * is in progress or when the iterators generated by {@link #children} are advanced.
   *
   * @deprecated Use {@link com.google.common.graph.Traverser#depthFirstPreOrder} instead, which has
   *     the same behavior.
   */

   *     value.)
   */
  @CanIgnoreReturnValue
  @CheckForNull
  V forcePut(@ParametricNullness K key, @ParametricNullness V value);

  // Bulk Operations

  /**
   * {@inheritDoc}
   *
   * <p><b>Warning:</b> the results of calling this method may vary depending on the iteration order
   * of {@code map}.
   *
   * @throws IllegalArgumentException if an attempt to {@code put} any entry fails. Note that some

- MinIO object storage set up per <https://min.io/docs/minio/linux/index.html>
- Veeam requires TLS connections to the object storage.  This can be configured per <https://min.io/docs/minio/linux/operations/network-encryption.html>
- The S3 bucket, Access Key and Secret Key have to be created before and outside of Veeam.

                this.specs = uncheckedCast(Iterators.toArray(iterator, Spec.class));
            }
        }
    }

    @SuppressWarnings({"unchecked", "TypeParameterUnusedInFormals"})
    static <T> T uncheckedCast(Object object) {
        return (T) object;
    }

    // Not public. Evaluation of these specs is a major hot spot for large builds, so use an array for iteration
    Spec<? super T>[] getSpecsArray() {

// Picks the device for which XLA should compile a cluster that contains
// operations placed in devices in `devices`.  For instance a cluster that
// contains operations solely placed on the CPU will be compiled into a CPU
// executable by XLA, whereas a cluster that contains operations placed on the
// CPU and also operations placed on the GPU will be compiled into a GPU
// executable.
//

            if (graph.isDirected()) {
              return Iterators.unmodifiableIterator(
                  Iterators.concat(
                      Iterators.transform(
                          graph.predecessors(node).iterator(),
                          (N predecessor) -> EndpointPair.ordered(predecessor, node)),
                      Iterators.transform(

            if (graph.isDirected()) {
              return Iterators.unmodifiableIterator(
                  Iterators.concat(
                      Iterators.transform(
                          graph.predecessors(node).iterator(),
                          (N predecessor) -> EndpointPair.ordered(predecessor, node)),
                      Iterators.transform(

/**
 * CompactLinkedHashSet is an implementation of a Set, which a predictable iteration order that
 * matches the insertion order. All optional operations (adding and removing) are supported. All
 * elements, including {@code null}, are permitted.
 *
 * <p>{@code contains(x)}, {@code add(x)} and {@code remove(x)}, are all (expected and amortized)
 * constant time operations. Expected in the hashtable sense (depends on the hash function doing a

/**
 * CompactLinkedHashSet is an implementation of a Set, which a predictable iteration order that
 * matches the insertion order. All optional operations (adding and removing) are supported. All
 * elements, including {@code null}, are permitted.
 *
 * <p>{@code contains(x)}, {@code add(x)} and {@code remove(x)}, are all (expected and amortized)
 * constant time operations. Expected in the hashtable sense (depends on the hash function doing a