* allow restricting subresource access ([#29988](https://github.com/kubernetes/kubernetes/pull/29988), [@deads2k](https://github.com/deads2k))
* Add density (batch pods creation latency and resource) and resource performance tests to `test-e2e-node' ([#29764](https://github.com/kubernetes/kubernetes/pull/29764), [@coufon](https://github.com/coufon))

      --project "${NETWORK_PROJECT}" \
      --network "${NETWORK}" \
      --source-ranges "0.0.0.0/0" \
      --allow "tcp:22" &
  fi

  # Open up TCP 3389 to allow RDP connections.
  if [[ ${NUM_WINDOWS_NODES} -gt 0 ]]; then
    if ! gcloud compute firewall-rules describe --project "${NETWORK_PROJECT}" "${NETWORK}-default-rdp" &>/dev/null; then

    server.enqueue(
      MockResponse.Builder()
        .addHeader("Allow: GET, HEAD")
        .addHeader("Last-Modified: " + formatDate(-1, TimeUnit.HOURS))
        .addHeader("Cache-Control: max-age=0")
        .body("A")
        .build(),
    )
    server.enqueue(
      MockResponse.Builder()
        .addHeader("Allow: GET, HEAD, PUT")
        .code(HttpURLConnection.HTTP_NOT_MODIFIED)

				"self.emptyStrings.max() == 'c'": "max called on empty list",

				// only allow sum on numeric types and duration
				"['a', 'b'].sum() == 'c'": "found no matching overload for 'sum' applied to 'list(string).()", // compiler type checking error

				// only allow min/max/indexOf/lastIndexOf on comparable types

* Fix panic in vSphere cloud provider ([#38423](https://github.com/kubernetes/kubernetes/pull/38423), [@BaluDontu](https://github.com/BaluDontu))
* Allow missing keys in templates by default ([#39486](https://github.com/kubernetes/kubernetes/pull/39486), [@ncdc](https://github.com/ncdc))

  let constructor = "tf_executor::CreateTFExecutorTPUV1IslandOutliningPass()";
  let description = [{
    Extract the islands containing a TPU cluster computation into an outlined
    function in a nested module. This will allow to run the usual bridge on this
    nested module which now exhibits a more friendly "V2-like" structure.
    This is only intended for V1 compatibility mode where the bridge runs without
    feed/fetches on session create/extend.

}

StringRef BiasAddOp::GetOptimalLayout(const RuntimeDevices& devices) {
  // Keep current data format if no GPUs are available or if explicit placement
  // does not allow to use GPU for this operation.
  if (!CanUseGpuDevice(devices) || !CanUseGpuDevice(getOperation()))
    return getDataFormat();

  // Prefer NHWC for GPU devices.
  return "NHWC";
}

                This is much simpler (less smart) than `response_model_exclude_unset`
                and `response_model_exclude_defaults`. You probably want to use one of
                those two instead of this one, as those allow returning `None` values
                when it makes sense.

                Read more about it in the

								Build(),
						},
						Count: 1,
						Check: check.And(
							check.OK(),
							check.ResponseHeaders(map[string]string{
								"Access-Control-Allow-Origin":  "cors.com",
								"Access-Control-Allow-Methods": "POST,GET",
								"Access-Control-Allow-Headers": "X-Foo-Bar,X-Foo-Baz",
								"Access-Control-Max-Age":       "86400",
							})),
					}
				}(),
			},
			{
				name: "get",

                This is much simpler (less smart) than `response_model_exclude_unset`
                and `response_model_exclude_defaults`. You probably want to use one of
                those two instead of this one, as those allow returning `None` values
                when it makes sense.

                Read more about it in the