* Fix GPU resource validation ([#28743](https://github.com/kubernetes/kubernetes/pull/28743), [@therc](https://github.com/therc))
* Scale kube-proxy conntrack limits by cores (new default behavior) ([#28876](https://github.com/kubernetes/kubernetes/pull/28876), [@thockin](https://github.com/thockin))

- Fixed #118052: nodeAffinity on pods can change prior to scheduling gates being removed even when podSpec.affinity is nil in the initial spec, this matches the 1.28 behavior to allow consistent integrators to be written. ([#118199](https://github.com/kubernetes/kubernetes/pull/118199), [@aleskandro](https://github.com/aleskandro)) [SIG Apps, Scheduling and Testing]

                 * child's <relativePath> should point at that parent, too. If it doesn't, we ignore the cache and
                 * resolve externally, to mimic the behavior if the cache didn't exist in the first place. Otherwise,
                 * the cache would obscure a bad POM.
                 */

                File pomFile = parentData.getModel().getPomFile();

- Prevent watch cache starvation by moving its watch to separate RPC and add a SeparateCacheWatchRPC feature flag to disable this behavior ([#123694](https://github.com/kubernetes/kubernetes/pull/123694), [@mengqiy](https://github.com/mengqiy)) [SIG API Machinery]

- Reverted a kubectl azure auth module change where oidc claim spn: prefix was omitted resulting a breaking behavior with existing Azure AD OIDC enabled api-server ([#87507](https://github.com/kubernetes/kubernetes/pull/87507), [@weinong](https://github.com/weinong)) [SIG API Machinery, Auth and Cloud Provider]

pkg syscall (windows-amd64), func LoadCreateSymbolicLink() error

# CL 149510043 net: disable SIO_UDP_CONNRESET behavior on windows., Ron Hashimoto <******@****.***>
pkg syscall (windows-386), const SIO_UDP_CONNRESET = 2550136844
pkg syscall (windows-386), const SIO_UDP_CONNRESET ideal-int

> open connections to the database. This setting is generally NOT recommended as the behavior may be inconsistent during recursive deletes in `namespace` format.

        } else if (commandLine.hasOption(CLIManager.FAIL_NEVER)) {
            return MavenExecutionRequest.REACTOR_FAIL_NEVER;
        } else {
            // this is the default behavior.
            return MavenExecutionRequest.REACTOR_FAIL_FAST;
        }
    }

    private TransferListener determineTransferListener(
            final boolean quiet,
            final boolean verbose,

common:windows_x86_cpu_2022 --platforms="//tensorflow/tools/toolchains/win2022:windows_ltsc2022_clang"

# Options to build TensorFlow 1.x or 2.x.
# TODO(kanglan): Change v2's define to default behavior
common:v2 --define=tf_api_version=2 --action_env=TF2_BEHAVIOR=1

# Enable all targets in XLA
common:cpu_cross --define=with_cross_compiler_support=true

# Disable XLA on mobile.

implementation is changed to not return an error when the specified target can be detected as not a mount point. On Linux, the behavior of detecting a mount point depends on `umount` command is validated when the mounter is created. Additionally, mount point checks will be skipped in CleanupMountPoint/CleanupMountWithForce if the mounter's Unmount having the changed behavior of not returning error when target is not a mount point. ([#109676](https://github.com/kubernetes/kubernetes/pull/109676), [...