* com.google.common.io.ByteStreams#newDataOutput()} to get a growable buffer.
   */
  public static byte[] toByteArray(long value) {
    // Note that this code needs to stay compatible with GWT, which has known
    // bugs when narrowing byte casts of long values occur.
    byte[] result = new byte[8];
    for (int i = 7; i >= 0; i--) {
      result[i] = (byte) (value & 0xffL);
      value >>= 8;
    }
    return result;
  }

 * loading), and fewer <a
 * href="https://github.com/google/guava/issues?q=is%3Aopen+is%3Aissue+label%3Apackage%3Dcache+label%3Atype%3Ddefect">bugs</a>.
 *
 * <p>Caffeine defines its own interfaces (<a
 * href="https://www.javadoc.io/doc/com.github.ben-manes.caffeine/caffeine/latest/com.github.benmanes.caffeine/com/github/benmanes/caffeine/cache/Cache.html">{@code

   *     ImmutableSortedSet#copyOf(Comparable[])}.</b>
   */
  @DoNotCall("Pass parameters of type Comparable")
  @Deprecated
  // The usage of "Z" here works around bugs in Javadoc (JDK-8318093) and JDiff.
  public static <Z> ImmutableSortedSet<Z> copyOf(Z[] elements) {
    throw new UnsupportedOperationException();
  }

 *
 * <p>This version is the fewest lines of code, largely because the synchronization mechanism used
 * is built into the language and runtime. But the programmer has to remember to avoid a couple of
 * common bugs: The {@code wait()} must be inside a {@code while} instead of an {@code if}, and
 * {@code notifyAll()} must be used instead of {@code notify()} because there are two different
 * logical conditions being awaited.
 *

   *     the {@link ExecutionException} thrown by {@link Future#get()} on the {@link Future}
   *     underlying this step or, if {@code get()} throws a different kind of exception, that
   *     exception itself. To avoid hiding bugs and other unrecoverable errors, callers should
   *     prefer more specific types, avoiding {@code Throwable.class} in particular.
   * @param fallback the function to be called if this step fails with the expected exception type.

        TF_NewWhile(func_graph_, &inputs[0], inputs.size(), s_)));
    ASSERT_EQ(TF_OK, TF_GetCode(s_)) << TF_Message(s_);
    params->name = "test_loop";

    // Initialize outputs so we can easily detect errors/bugs
    outputs.resize(2, {nullptr, -1});

    // Create loop: while (input1 < input2) input1 += input2 + 1
    TF_Operation* less_than = LessThan(

    harmed connection reuse.
 *  Fix: Ensure canceled and discarded HTTP/2 data is not permanently counted against the limited
    flow control window. We had a few bugs where window size accounting was broken when streams
    were canceled or reset.
 *  Fix: Recover gracefully if the TLS session returns an unexpected version (`NONE`) or cipher
    suite (`SSL_NULL_WITH_NULL_NULL`).

        *   The new backend is used for 8 bits full integer post-training
            quantization
        *   The new backend removes the redundant rescales and fixes some bugs
            (shared weight/bias, extremely small scales, etc)
        *   Set `experimental_new_quantizer` in tf.lite.TFLiteConverter to False
            to disable this change
*   `tf.keras`

### SIG-OpenStack

    assertThat(cache.hitCount()).isEqualTo(2)
    assertThat(response2.handshake!!.cipherSuite).isEqualTo(
      response1.handshake!!.cipherSuite,
    )
  }

  /**
   * We've had bugs where caching and cross-protocol redirects yield class cast exceptions internal
   * to the cache because we incorrectly assumed that HttpsURLConnection was always HTTPS and