- Fixed issue on Windows when calculating cpu limits on nodes with more than 64 logical processors ([#114231](https://github.com/kubernetes/kubernetes/pull/114231), [@mweibel](https://github.com/mweibel))

27CD          ; valid                  ;      ; NV8    # 6.1  MATHEMATICAL FALLING DIAGONAL
27CE..27CF    ; valid                  ;      ; NV8    # 6.0  SQUARED LOGICAL AND..SQUARED LOGICAL OR
27D0..27EB    ; valid                  ;      ; NV8    # 3.2  WHITE DIAMOND WITH CENTRED DOT..MATHEMATICAL RIGHT DOUBLE ANGLE BRACKET

  clusterrolebindings, this command computes covers and adds the missing rules.
  The logic required to properly apply RBAC permissions is more complicated
  than a JSON merge because you have to compute logical covers operations between
  rule sets. This means that we cannot use `kubectl apply` to update RBAC roles
  without risking breaking old clients, such as controllers.

          "type": "string"
        }
      },
      "type": "object"
    },
    "io.k8s.api.discovery.v1.Endpoint": {
      "description": "Endpoint represents a single logical \"backend\" implementing a service.",
      "properties": {
        "addresses": {

            is not there, it means they are off.

## Bug Fixes and Other Changes

*   New argument `experimental_device_ordinal` in `LogicalDeviceConfiguration`
    to control the order of logical devices (GPU only).

*   `tf.keras`:

    *   Changed the TensorBoard tag names produced by the
        `tf.keras.callbacks.TensorBoard` callback, so that summaries logged

          "type": "string"
        }
      },
      "type": "object"
    },
    "io.k8s.api.discovery.v1.Endpoint": {
      "description": "Endpoint represents a single logical \"backend\" implementing a service.",
      "properties": {
        "addresses": {