- Pod will not start when Init container was OOM killed. ([#104650](https://github.com/kubernetes/kubernetes/pull/104650), [@yxxhero](https://github.com/yxxhero)) [SIG Node]

- If using cgroups v2, then the cgroup aware OOM killer will be enabled for container cgroups via  `memory.oom.group` .  This causes processes within the cgroup to be treated as a unit and killed simultaneously in the event of an OOM kill on any process in the cgroup. ([#117793](https://github.com/kubernetes/kubernetes/pull/117793), [@tzneal](https://github.com/tzneal)) [SIG Apps, Node and Testing]...

* Fix API Server potential memory leak issue in processing watch request. ([#85410](https://github.com/kubernetes/kubernetes/pull/85410), [@answer1991](https://github.com/answer1991))
* Verify kubelet & kube-proxy can recover after being killed on Windows nodes ([#84886](https://github.com/kubernetes/kubernetes/pull/84886), [@YangLu1031](https://github.com/YangLu1031))

- GCE/Windows: When the service cannot be stopped Stackdriver logging processes are now force killed ([#77378](https://github.com/kubernetes/kubernetes/pull/77378), [@yujuhong](https://github.com/yujuhong))

- Change the calculation of pod UIDs so that static pods get a unique value - will cause all containers to be killed and recreated after in-place upgrade. ([#87461](https://github.com/kubernetes/kubernetes/pull/87461), [@bboreham](https://github.com/bboreham)) [SIG Node]

    return super.pendingToString();
  }

  /**
   * Must be called at the end of each subclass's constructor. This method performs the "real"
   * initialization; we can't put this in the constructor because, in the case where futures are
   * already complete, we would not initialize the subclass before calling {@link
   * #collectValueFromNonCancelledFuture}. As this is called after the subclass is constructed,

				// The done() might have been called
				// concurrently, check for it before we
				// write the filler byte.
				select {
				case err := <-doneCh:
					if err != nil {
						write([]byte{1})
						write([]byte(err.Error()))
					} else {
						write([]byte{0})
					}
					return
				default:
				}

				// Response not ready, write a filler byte.
				write([]byte{32})

      RegularImmutableMap<K, V> empty = (RegularImmutableMap<K, V>) EMPTY;
      return empty;
    } else if (n == 1) {
      // requireNonNull is safe because the first `2*n` elements have been filled in.
      checkEntryNotNull(
          requireNonNull(alternatingKeysAndValues[0]), requireNonNull(alternatingKeysAndValues[1]));
      return new RegularImmutableMap<K, V>(null, alternatingKeysAndValues, 1);
    }

TensorFlow's official CI jobs run the scripts in this folder. Our internal CI
system, Kokoro, schedules our CI jobs by combining a build script with a file
from the `envs` directory that is filled with configuration options:

-   Nightly jobs (Run nightly on the `nightly` branch)
    -   Uses `wheel.sh`, `libtensorflow.sh`, `code_check_full.sh`
-   Continuous jobs (Run on every GitHub commit)
    -   Uses `pycpp.sh`

      switch (size) {
        case 0:
          return of();
        case 1:
          // requireNonNull is safe because the first `size` elements have been filled in.
          Entry<K, V> onlyEntry = requireNonNull(entries[0]);
          return of(onlyEntry.getKey(), onlyEntry.getValue());
        default:
          /*
           * If entries is full, or if hash flooding is detected, then this implementation may end