* Kubectl
  * Display line number on JSON errors
  * Add flag -t as shorthand for --tty
* Resources
  * Improved node stability by *optionally* evicting pods upon memory pressure - [Design Doc](https://github.com/kubernetes/kubernetes/blob/release-1.3/docs/proposals/kubelet-eviction.md)
  * **alpha**: NVIDIA GPU support ([#24836](https://github.com/kubernetes/kubernetes/pull/24836), [@therc](https://github.com/therc))

   * granularity. For example, access queues and write queues are kept per segment when they are
   * required by the selected eviction algorithm. As such, when writing unit tests it is not
   * uncommon to specify {@code concurrencyLevel(1)} in order to achieve more deterministic eviction
   * behavior.
   *
   * <p>Note that future implementations may abandon segment locking in favor of more advanced
   * concurrency controls.

                        dependency);
            }

            Dependency existing = index.get(key);

            if (existing != null) {
                String msg;
                if (equals(existing.getVersion(), dependency.getVersion())) {
                    msg = "duplicate declaration of version " + Objects.toString(dependency.getVersion(), "(?)");

We will now enable bucket event notifications on a bucket named `images`. Whenever a JPEG image is created/overwritten, a new document is added or an existing document is updated in the Elasticsearch index configured above. When an existing object is deleted, the corresponding document is deleted from the index. Thus, the rows in the Elasticsearch index, reflect the `.jpg` objects in the `images` bucket.

// -1 and `dims` can be null. If a dimension is unknown, the
// corresponding entry in the `dims` array must be -1.
//
// This does not overwrite the existing shape associated with `output`,
// but merges the input shape with the existing shape.  For example,
// setting a shape of [-1, 2] with an existing shape [2, -1] would set
// a final shape of [2, 2] based on shape merging semantics.
//
// Returns an error into `status` if:

   * granularity. For example, access queues and write queues are kept per segment when they are
   * required by the selected eviction algorithm. As such, when writing unit tests it is not
   * uncommon to specify {@code concurrencyLevel(1)} in order to achieve more deterministic eviction
   * behavior.
   *
   * <p>Note that future implementations may abandon segment locking in favor of more advanced
   * concurrency controls.

		// Test Case - 7.
		// Existing bucket, but using a bucket on which NewMultipartUpload is not Initiated.
		{bucketName: "unused-bucket", objName: "def", uploadID: "xyz", PartID: 1, expectedError: fmt.Errorf("%s", "Invalid upload id xyz")},
		// Test Case - 8.
		// Existing bucket, object name different from which NewMultipartUpload is constructed from.

		expectedErr     error
	}{
		// TestXLStorage case - 1.
		// valid case with existing volume and file to delete.
		{
			srcVol:          "success-vol",
			srcPath:         "abc",
			expectedListDir: []string{"def/", "xyz/"},
			expectedErr:     nil,
		},
		// TestXLStorage case - 1.
		// valid case with existing volume and file to delete.
		{
			srcVol:          "success-vol",

		cache.iamUserGroupMemberships = newCache.iamUserGroupMemberships
		cache.iamUserPolicyMap = newCache.iamUserPolicyMap
		cache.iamUsersMap = newCache.iamUsersMap
		// For STS policy map, we need to merge the new cache with the existing
		// cache because the periodic IAM reload is partial. The periodic load
		// here is to account for STS policy mapping changes that should apply
		// for service accounts derived from such STS accounts (i.e. LDAP STS

	x.versions = append(x.versions, xlMetaV2ShallowVersion{header: xlMetaV2VersionHeader{ModTime: -1}})

	// Linear search, we likely have to insert at front.
	for i, existing := range x.versions {
		if existing.header.ModTime <= modTime {
			// Insert at current idx. First move current back.
			copy(x.versions[i+1:], x.versions[i:])
			x.versions[i] = xlMetaV2ShallowVersion{
				header: ver.header(),