* Fix incorrectly propagated ResourceVersion in ListRequests returning 0 items. ([#64150](https://github.com/kubernetes/kubernetes/pull/64150), [@wojtek-t](https://github.com/wojtek-t))

protobuf you will be forced to restore from backup, and you will lose any changes
since you converted to protobuf.  After conversion to protobuf, you should
validate the correct operation of your cluster thoroughly before returning it
to normal operation.

Backups are always a good precaution, particularly around upgrades, but this
upgrade has multiple known issues where the only remedy is to restore from
backup.

* Protobuf serialization does not distinguish between `[]` and `null`. ([#45294](https://github.com/kubernetes/kubernetes/pull/45294), [@liggitt](https://github.com/liggitt))

- Kube-apiserver: fixed a bug returning inconsistent results from list requests which set a field or label selector and set a paging limit ([#94002](https://github.com/kubernetes/kubernetes/pull/94002), [@wojtek-t](https://github.com/wojtek-t)) [SIG API Machinery]

- Kube-apiserver: fixed a bug returning inconsistent results from list requests which set a field or label selector and set a paging limit ([#94002](https://github.com/kubernetes/kubernetes/pull/94002), [@wojtek-t](https://github.com/wojtek-t)) [SIG API Machinery]

      /*
       * requireNonNull is safe because propertyNames contains only non-null elements.
       *
       * Accordingly, we have it annotated as returning `Enumeration<? extends Object>` in our
       * prototype checker's JDK. However, the checker still sees the return type as plain
       * `Enumeration<?>`, probably because of one of the following two bugs (and maybe those two

      /*
       * requireNonNull is safe because propertyNames contains only non-null elements.
       *
       * Accordingly, we have it annotated as returning `Enumeration<? extends Object>` in our
       * prototype checker's JDK. However, the checker still sees the return type as plain
       * `Enumeration<?>`, probably because of one of the following two bugs (and maybe those two

        `tf.keras.utils.image_dataset_from_directory`,`tf.keras.utils.text_dataset_from_directory`,
        and `audio_dataset_from_directory`, to be used with the
        `validation_split` argument, for returning both dataset splits at once,
        as a tuple.
    *   Added `tf.keras.utils.split_dataset` utility to split a `Dataset` object
        or a list/tuple of arrays into two `Dataset` objects (e.g. train/test).

Type inference supports calls of generic functions and assignments
of generic functions to (explicitly function-typed) variables.
This includes passing generic functions as arguments to other
(possibly also generic) functions, and returning generic functions
as results.
Type inference operates on a set of equations specific to each of
these cases.
The equations are as follows (type argument lists are omitted for clarity):
</p>

<ul>

* Filter out non-RestartAlways mirror pod in restart test. ([#37203](https://github.com/kubernetes/kubernetes/pull/37203), [@Random-Liu](https://github.com/Random-Liu))
* Validate volume spec before returning azure mounter ([#37018](https://github.com/kubernetes/kubernetes/pull/37018), [@rootfs](https://github.com/rootfs))