TF's boosted trees code will be eliminated before the branch cut for TF 2.9
    and will no longer be present since that release.

## Security

*   Fixes a floating point division by 0 when executing convolution operators
    ([CVE-2022-21725](https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2022-21725))
*   Fixes a heap OOB read in shape inference for `ReverseSequence`

  - apiserver_flowcontrol_latest_s: S(the request last dispatched) = R(when that request starts executing in the virtual world)
  - apiserver_flowcontrol_next_s_bounds: min and max next S among non-empty queues

     */
    void postReadCleanup() {
      if ((readCount.incrementAndGet() & DRAIN_THRESHOLD) == 0) {
        runCleanup();
      }
    }

    /**
     * Performs routine cleanup prior to executing a write. This should be called every time a write
     * thread acquires the segment lock, immediately after acquiring the lock.
     */
    @GuardedBy("this")
    void preWriteCleanup() {
      runLockedCleanup();

`ClusterRole` access. The User in `super-admin.conf` is now bound to the `system:masters` built-in super-powers / break-glass Group that can bypass RBAC. Before this change, the default `admin.conf` was bound to `system:masters` Group, which was undesired. Executing `kubeadm init phase kubeconfig all` or just `kubeadm init` will now generate the new `super-admin.conf` file. The cluster admin can then decide to keep the file present on a node host or move it to a safe location. `kubadm certs renew` will renew...

  - **apiserver_flowcontrol_priority_level_seat_count_samples**: histograms of seats occupied by executing requests (both regular and final-delay phases included), broken down by priority_level; the observations are taken once per millisecond.
  - **apiserver_flowcontrol_priority_level_seat_count_watermarks**: histograms of high and low watermarks of number of seats occupied by executing requests (both regular and final-delay phases included), broken down by priority_level.

- The apiserver debug endpoint `/debug/api_priority_and_fairness/dump_requests` has been extended to dump executing requests as well as queued ones.  A column for StartTime has been added to the returned table, with the queued requests having a StartTime of "0001-01-01T00:00:00Z".  The executing requests have a RequestIndexInQueue of -1, and the QueueIndex is also -1 for priority levels without queues. ([#119009](https://github.com/kubernetes/kube...

- kubeadm: fixed a misleading output (typo) about control-plane joining instructions when executing the "kubeadm init" command. ([#128118](https://github.com/kubernetes/kubernetes/pull/128118), [@amaddio](https://github.com/amaddio))

- Kubeadm: don't write the kubelet environment file on "upgrade apply" ([#85412](https://github.com/kubernetes/kubernetes/pull/85412), [@boluisa](https://github.com/boluisa)) [SIG Cluster Lifecycle]
- Kubeadm: fix potential panic when executing "kubeadm reset" with a corrupted kubelet.conf file ([#86216](https://github.com/kubernetes/kubernetes/pull/86216), [@neolit123](https://github.com/neolit123)) [SIG Cluster Lifecycle]

- Kubelet.exe on Windows now checks that the process running as administrator and the executing user account is listed in the built-in administrators group.  This is the equivalent to checking the process is running as uid 0. ([#96616](https://github.com/kubernetes/kubernetes/pull/96616), [@perithompson](https://github.com/perithompson)) [SIG...

on the Program.

  To "propagate" a work means to do anything with it that, without
permission, would make you directly or secondarily liable for
infringement under applicable copyright law, except executing it on a
computer or modifying a private copy.  Propagation includes copying,
distribution (with or without modification), making available to the
public, and in some countries other activities as well.