*     the {@link ExecutionException} thrown by {@code input.get()} 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 {@link Function} to be called if {@code input} fails with the expected

   * avoid memory leaks when an element is no longer necessary.
   *
   * <p>This method accepts an array with element type {@code @Nullable T}, but callers must pass an
   * array whose contents are initially non-null. The {@code @Nullable} annotation indicates that
   * this method will write nulls into the array during iteration.
   *

   * avoid memory leaks when an element is no longer necessary.
   *
   * <p>This method accepts an array with element type {@code @Nullable T}, but callers must pass an
   * array whose contents are initially non-null. The {@code @Nullable} annotation indicates that
   * this method will write nulls into the array during iteration.
   *

// versionID, NOTE: versionID must be valid and should exist -
// and must not be a DeleteMarker or legacy object, if no
// versionID is specified 'null' versionID is updated instead.
//
// It is callers responsibility to set correct versionID, this
// function shouldn't be further extended to update immutable
// values such as ErasureInfo, ChecksumInfo.
//
// Metadata is only updated to new values, existing values

      boolean expression,
      /*
       * TODO(cpovirk): Consider removing @CheckForNull here, as we've done with the other methods'
       * errorMessageTemplate parameters: It is unlikely that callers intend for their string
       * template to be null (though we do handle that case gracefully at runtime). I've left this
       * one as it is because one of our users has defined a wrapper API around Preconditions,

      boolean expression,
      /*
       * TODO(cpovirk): Consider removing @CheckForNull here, as we've done with the other methods'
       * errorMessageTemplate parameters: It is unlikely that callers intend for their string
       * template to be null (though we do handle that case gracefully at runtime). I've left this
       * one as it is because one of our users has defined a wrapper API around Preconditions,

  /// implemented by a filesystem registered to handle the `fs://` scheme.
  ///
  /// A new `char*` buffer must be allocated by this method. Core TensorFlow
  /// manages the lifetime of the buffer after the call. Thus, all callers of
  /// this method must take ownership of the returned pointer.
  ///
  /// The implementation should clean up paths, including but not limited to,
  /// removing duplicate `/`s, and resolving `..` and `.`.
  ///

      .isEqualTo("http://host/?a=%21%24%28%29%2C%2F%3A%3B%3F%40%5B%5D%5C%5E%60%7B%7C%7D%7E")
    assertThat(url.queryParameter("a")).isEqualTo("!$(),/:;?@[]\\^`{|}~")
  }

  /**
   * When callers use `addEncodedQueryParameter()` we only encode what's strictly required. We
   * retain the encoded (or non-encoded) state of the input.
   */
  @Test
  fun queryCharactersNotReencodedWhenComposedWithAddEncoded() {

                }
                else if ( last == null ) {
                    // don't have enough credits/space for the first request, block until available
                    // for space there is nothing we can do, callers need to make sure that a single message fits

                    try {
                        long timeout = getResponseTimeout(chain);
                        if ( params.contains(RequestParam.NO_TIMEOUT) ) {

// * int is used as an index into arrays.
// * Deletion functions are safe to call on nullptr.
//
// Questions left to address:
// * Might at some point need a way for callers to provide their own Env.
// * Maybe add TF_TensorShape that encapsulates dimension info.
//
// Design decisions made:
// * Backing store for tensor memory has an associated deallocation