const char* attr_name,
                                          TF_Status* status) {
  const tensorflow::AttrValue* attr = oper->node.attrs().Find(attr_name);
  if (attr == nullptr) {
    status->status = InvalidArgument("Operation '", oper->node.name(),
                                     "' has no attr named '", attr_name, "'.");
  }
  return attr;
}

    // the given node name (error_node_) and requested device (error_device_).
    for (const auto node : graph->get()->nodes()) {
      if (node->name().find(error_node_) != string::npos &&
          node->requested_device() == error_device_) {
        return tensorflow::errors::Internal("Injected graph pass error.");
      }
    }
    return absl::OkStatus();
  }

	// Before Istio would update the pod manifest to rewrite healthchecks to go to sidecar Envoy port 15021,
	// so that it could distinguish things that can be unauthenticated (healthchecks) from other kinds of node traffic
	// (e.g. LoadBalanced Service packets, etc) that need to be authenticated/captured/proxied.
	//
	// We want to do the same thing in ambient but can't rely on podSpec injection. So, do effectively the same thing,

    assertTrue(Files.exists(parent));
    assertTrue(Files.exists(grandparent));
  }

  public void testCreateParentDirectories_noPermission() {
    if (isWindows()) {
      return; // TODO: b/136041958 - Create/find a directory that we don't have permissions on?
    }
    Path file = root().resolve("parent/nonexistent.file");
    Path parent = file.getParent();
    assertFalse(Files.exists(parent));

then fixing the bug will break that code.
New features can also have similar impacts:
enabling the HTTP/2 use by the HTTP client broke programs
connecting to servers with buggy HTTP/2 implementations.
These kinds of changes are unavoidable and
[permitted by the Go 1 compatibility rules](/doc/go1compat).
Even so, Go provides a mechanism called GODEBUG to
reduce the impact such changes have on Go developers

    for (int i = 0; i < args.length; i++) {
      Parameter param = params.get(i);
      if (i != indexOfParamToSetToNull) {
        args[i] = getDefaultValue(param.getType());
        Assert.assertTrue(
            "Can't find or create a sample instance for type '"
                + param.getType()
                + "'; please provide one using NullPointerTester.setDefault()",
            args[i] != null || isNullable(param));
      }
    }

 *         .catching(RpcException.class, e -> false, directExecutor());
 * }</pre>
 *
 * <h3>Alternatives</h3>
 *
 * <h4>Frameworks</h4>
 *
 * <p>When chaining together a graph of asynchronous operations, you will often find it easier to
 * use a framework. Frameworks automate the process, often adding features like monitoring,
 * debugging, and cancellation. Examples of frameworks include:
 *
 * <ul>

    development to avoid installing the packages directly on your system (in
    which case remember to change the directory from `/root` to `/tensorflow`
    once you get into the running container so `bazel` can find the `tensorflow`
    workspace).

    you can do this by using the following command. As an example-

    ```bash
    docker run -it --rm -v $PWD:/tmp -w /tmp tensorflow/build:2.15-python3.10
    ```

   */
  private static MatchResult createMatchResult() {
    Matcher matcher = Pattern.compile(".").matcher("X");
    matcher.find();
    return matcher.toMatchResult();
  }

  private static final ClassToInstanceMap<Object> DEFAULTS =
      ImmutableClassToInstanceMap.builder()
          // primitives
          .put(Object.class, "")

 *       href="https://github.com/google/guava/wiki/LivingWithNullHostileCollections">several other
 *       approaches to this</a> that should be considered first)
 * </ul>
 *
 * <p>A common alternative to using this class is to find or create a suitable <a
 * href="http://en.wikipedia.org/wiki/Null_Object_pattern">null object</a> for the type in question.
 *
 * <p>This class is not intended as a direct analogue of any existing "option" or "maybe" construct