* higher value than you need can waste space and time, and a significantly lower value can lead
   * to thread contention. But overestimates and underestimates within an order of magnitude do not
   * usually have much noticeable impact. A value of one permits only one thread to modify the cache
   * at a time, but since read operations and cache loading computations can proceed concurrently,

First, we have a concept of a `Full` push. Only `Full` pushes will recompute `PushContext` on change; otherwise this is skipped and the last `PushContext` is re-used. Note: even when `Full`, we try to copy as much from the previous `PushContext` as possible. For example, if only a `WasmPlugin` changed, we would not recompute services indexes.

    assertThat((long) arraysDim1 * arraysDim2).isNotEqualTo((long) (arraysDim1 * arraysDim2));

    double[][] arrays = new double[arraysDim1][];
    // it's shared to avoid using too much memory in tests
    double[] sharedArray = new double[arraysDim2];
    Arrays.fill(arrays, sharedArray);

    try {
      Doubles.concat(arrays);
      fail();
    } catch (IllegalArgumentException expected) {

   * input byte array {@code {0x54, 0x32}} would yield the {@code char} value {@code '\\u5432'}.
   *
   * <p>Arguably, it's preferable to use {@link java.nio.ByteBuffer}; that library exposes much more
   * flexibility at little cost in readability.
   *
   * @throws IllegalArgumentException if {@code bytes} has fewer than 2 elements
   */
  @GwtIncompatible // doesn't work

   * input byte array {@code {0x54, 0x32}} would yield the {@code short} value {@code 0x5432}.
   *
   * <p>Arguably, it's preferable to use {@link java.nio.ByteBuffer}; that library exposes much more
   * flexibility at little cost in readability.
   *
   * @throws IllegalArgumentException if {@code bytes} has fewer than 2 elements
   */
  @GwtIncompatible // doesn't work

            }
            catch ( UnknownHostException uhe ) {
                /*
                 * Java cannot determine the localhost. This is basically a config
                 * issue on the host. There's not much we can do about it. Just
                 * to suppress NPEs that would result we can create a possibly bogus
                 * address. Pretty sure the below cannot actually thrown a UHE tho.
                 */

 * order to {@code [key2, key1]}. The {@link #entries()} iterator returns mutable map entries, and
 * {@link #replaceValues} attempts to preserve iteration order as much as possible.
 *
 * <p>The collections returned by {@link #keySet()} and {@link #asMap} iterate through the keys in
 * the order they were first added to the multimap. Similarly, {@link #get}, {@link #removeAll}, and

   * input byte array {@code {0x54, 0x32}} would yield the {@code short} value {@code 0x5432}.
   *
   * <p>Arguably, it's preferable to use {@link java.nio.ByteBuffer}; that library exposes much more
   * flexibility at little cost in readability.
   *
   * @throws IllegalArgumentException if {@code bytes} has fewer than 2 elements
   */
  @GwtIncompatible // doesn't work

```

TensorFlow kernels and TensorFlow's dependencies are still not built with
debugging information with `--config=dbg`, as issues occur on Linux if
there is too much debug info (see [this GitHub
issue](https://github.com/tensorflow/tensorflow/issues/48919) for context). If
you want to debug a kernel, you can compile specific files with `-g` using the

    assertThrows(IndexOutOfBoundsException.class, () -> iia3.subArray(1, 4));
  }

  /*
   * Whenever an implementation uses `instanceof` on a parameter instance, the test has to know that
   * (so much for "black box") and try instances that both do and don't pass the check. The "don't"
   * half of that is more awkward to arrange...
   */
  private static <T> Iterable<T> iterable(final Collection<T> collection) {