* The TensorBoard (TB) dependency has been removed starting with TF 2.21.

### Major Features and Improvements

* `tf.lite`
    * Adds int8 and int16x8 support for SQRT operator.
    * Adds int16x8 support for EQUAL and NOT_EQUAL operators.
    * Adds support for int2 type.
    * Adds support for int2/int4 in tfl.cast .
    * Adds support for SRQ int2 in tfl.fully_connected.
    * Adds support for int4 in tfl.slice.

* Increase allowed lag for ssh key sync loop in tunneler to allow for one failure ([#60068](https://github.com/kubernetes/kubernetes/pull/60068), [@wojtek-t](https://github.com/wojtek-t))
* Fixing a bug in OpenStack cloud provider, where dual stack deployments (IPv4 and IPv6) did not work well when using kubenet as the network plugin. ([#59749](https://github.com/kubernetes/kubernetes/pull/59749), [@zioproto](https://github.com/zioproto))

 Extended resources cannot be overcommitted, so make sure that request and limit are equal
 if both are present in a container spec.

 - The default Bootstrap Token created with `kubeadm init` v1.8 expires
 and is deleted after 24 hours by default to limit the exposure of the

- Fixed an issue in the Windows kube-proxy (winkernel) where IPv4 and IPv6 Service load balancers could be incorrectly shared, causing broken dual-stack Service behavior. The kube-proxy now tracks load balancers per IP family, enabling correct support for PreferDualStack and RequireDualStack Services on Windows nodes. ([#136374](https://github.com/kubernetes/kubernetes/pull/136374), [@prince...

- Fixed an issue in the Windows kube-proxy (winkernel) where IPv4 and IPv6 Service load balancers could be incorrectly shared, causing broken dual-stack Service behavior. The kube-proxy now tracks load balancers per IP family, enabling correct support for PreferDualStack and RequireDualStack Services on Windows nodes. ([#136373](https://github.com/kubernetes/kubernetes/pull/136373), [@prince...

- KEP-1880: Users of the new feature to add multiple service CIDR will use by default a dual-write strategy on the new ClusterIP allocators to avoid the problem of possible duplicate IPs allocated to Services when running skewed kube-apiservers using different allocators. They can opt-out of this behavior by enabled the feature gate Disab...

  
  When `LimitedSwap` is enabled the swap limit would be automatically calculated for
  Burstable QoS pods. For Best-Effort/Guaranteed QoS pods, swap would be disabled.
  
  Containers with memory requests equal to their memory limits also won't have
  swap access, and it is a way to opt-out of swap for a single container.
  
  The formula for the swap limit for Burstable QoS pods is:

    assertThat(server.takeRequest().exchangeIndex).isEqualTo(2)
  }

  /**
   * Each OkHttpClient used to get its own instance of NullProxySelector, and because these weren't
   * equal their connections weren't pooled. That's a nasty performance bug!
   *
   * https://github.com/square/okhttp/issues/5519
   */
  @Test
  fun connectionPoolingWithFreshClientSamePool() {

AAe,IACvB/mB,KAAM,CAAEgnB,QAAS/H,EAAEgI,cAAe5mC,IAAK4+B,EAAEgC,MACzCrb,GAAI,aAAc3iB,EAAW,SAAUikC,GACvCpmC,EAAOka,SACP/X,EAAW,KACNikC,GACJ5N,EAAuB,UAAb4N,EAAI9mC,KAAmB,IAAM,IAAK8mC,EAAI9mC,QAKnD/B,EAAS8C,KAAKC,YAAaN,EAAQ,KAEpC2iC,MAAO,WACDxgC,GACJA,QAUL,IAqGKigB,GArGDikB,GAAe,GAClBC,GAAS,oBAGV1lC,GAAOqgC,UAAW,CACjBsF,MAAO,WACPC,cAAe,WACd,IAAIrkC,EAAWkkC,GAAargC,OAAWpF,GAAOiD,QAAU,IAAQrE,GAAMmG,OAEtE,OADAhI,KAAMwE,IAAa,EACZA,KAKTvB,GAAOugC,cAAe,aAAc,SAAUhD,EAAGsI,EAAkBlH,GAElE,IAAImH,EAAcC,EAAaC,E...

1DFC          ; valid                                  # 6.0  COMBINING DOUBLE INVERTED BREVE BELOW
1DFD          ; valid                                  # 5.2  COMBINING ALMOST EQUAL TO BELOW
1DFE..1DFF    ; valid                                  # 5.0  COMBINING LEFT ARROWHEAD ABOVE..COMBINING RIGHT ARROWHEAD AND DOWN ARROWHEAD BELOW