Detect zstd:chunked format in source blobs

... instead of only treating it as zstd.

Signed-off-by: Miloslav Trmač <mitr@redhat.com>

copy/compression.go

@@ -52,6 +52,16 @@ func blobPipelineDetectCompressionStep(stream *sourceStream, srcInfo types.BlobI
 	}
 	stream.reader = reader
 
+	if decompressor != nil && format.Name() == compressiontypes.ZstdAlgorithmName {
+		tocDigest, err := chunkedToc.GetTOCDigest(srcInfo.Annotations)
+		if err != nil {
+			return bpDetectCompressionStepData{}, err
+		}
+		if tocDigest != nil {
+			format = compression.ZstdChunked
+		}
+
+	}
 	res := bpDetectCompressionStepData{
 		isCompressed: decompressor != nil,
 		format:       format,

copy/single.go

@@ -163,7 +163,7 @@ func (c *copier) copySingleImage(ctx context.Context, unparsedImage *image.Unpar
 		if format == nil {
 			format = defaultCompressionFormat
 		}
-		if format.Name() == compression.ZstdChunked.Name() {
+		if format.Name() == compressiontypes.ZstdChunkedAlgorithmName {
 			if ic.requireCompressionFormatMatch {
 				return copySingleImageResult{}, errors.New("explicitly requested to combine zstd:chunked with encryption, which is not beneficial; use plain zstd instead")
 			}