}
	return 0, ErrNotImplemented
}

func (lrw *ResponseRecorder) Write(p []byte) (int, error) {
	if !lrw.headersLogged {
		// We assume the response code to be '200 OK' when WriteHeader() is not called,
		// that way following Golang HTTP response behavior.
		lrw.WriteHeader(http.StatusOK)
	}
	n, err := lrw.ResponseWriter.Write(p)
	lrw.bytesWritten += n
	if lrw.ttfbBody == 0 {
		lrw.ttfbBody = time.Now().UTC().Sub(lrw.StartTime)

        writeLittleEndianInt(baos, PacConstants.LOGON_INFO); // type
        writeLittleEndianInt(baos, 100); // size
        writeLittleEndianLong(baos, 1000); // offset way out of bounds
        byte[] pacData = baos.toByteArray();

        // This currently throws ArrayIndexOutOfBoundsException
        // but should be wrapped in PACDecodingException

again.

[individual CLA]: https://cla.developers.google.com/about/google-individual
[corporate CLA]: https://cla.developers.google.com/about/google-corporate

## Building and Testing

The most reliable way to build and test Guava is:

```shell
./mvnw clean install
```

If you want to run a single test, you'll need to pass multiple flags because
Guava is a multi-module project. For performance reasons, you may also wish to

        that actually comes up in the real world. We want to hear the real-world use case so the
        community can discuss and debate whether this feature is actually the *best* way to address
        the real use case, or whether or not a different abstraction might be more appropriate.


        It's okay if you can't provide complete context on a use case. We understand if you are not

    }
  }

  /**
   * Wraps a {@code PrimitiveSink} as an {@link OutputStream}, so it is easy to {@link Funnel#funnel
   * funnel} an object to a {@code PrimitiveSink} if there is already a way to write the contents of
   * the object to an {@code OutputStream}.
   *
   * <p>The {@code close} and {@code flush} methods of the returned {@code OutputStream} do nothing,
   * and no method throws {@code IOException}.

      assertWrapsInterruptedException(expected);
    } finally {
      interruptenator.shutdown();
      Thread.interrupted();
    }
  }

  /**
   * awaitFullGc() is not quite as reliable a way to ensure calling of a specific finalize method as
   * the more direct await* methods, but should be reliable enough in practice to avoid flakiness of
   * this test. (And if it isn't, we'd like to know about it first!)
   */

// of default zeros grads is handled by the `DefaultGradientFunction` registered
// for each op.
// TODO(srbs): We need to define `ZerosLike` here to keep the compiler happy.
// Figure out a way to avoid this.
// TODO(srbs): Should ZerosLike check-fail instead of returning nullptr?
class TapeTensor {
 public:
  explicit TapeTensor(AbstractTensorHandle* handle);
  TapeTensor(const TapeTensor& other);

You could also create a `JSONResponse` directly and return it.

/// tip

You will normally have much better performance using a [Response Model](../tutorial/response-model.md) than returning a `JSONResponse` directly, as that way it serializes the data using Pydantic, in Rust.

///

## Return a `Response` { #return-a-response }

You can return a `Response` or any sub-class of it.

/// info

## `fastapi run` { #fastapi-run }

Executing `fastapi run` starts FastAPI in production mode.

By default, **auto-reload** is disabled. It also listens on the IP address `0.0.0.0`, which means all the available IP addresses, this way it will be publicly accessible to anyone that can communicate with the machine. This is how you would normally run it in production, for example, in a container.

   * }
   * }
   *
   * @param <X> The parameter type
   * @param typeParam the parameter type variable
   * @param typeArg the actual type to substitute
   */
  /*
   * TODO(cpovirk): Is there any way for us to support TypeParameter instances for type parameters
   * that have nullable bounds? Unfortunately, if we change the parameter to TypeParameter<? extends