fix: prevent panic in ParseCopySource on empty input

ParseCopySource indexes into copySourceHeader[0] without checking
for an empty string, causing an index-out-of-range panic.

The three backend callers (azure, posix) pass the x-amz-copy-source
header value directly, so a malformed or missing header propagates
an empty string into ParseCopySource.

Add an empty-string guard at the top of the function that returns
an InvalidArgCopySourceBucket error, consistent with the existing
error returned when the source path has no bucket/object separator.

Add a table-driven test case that reproduces the panic without the
fix and verifies the correct error with the fix.

Signed-off-by: Sebastien Tardif <sebtardif@ncf.ca>

backend/common.go

@@ -230,6 +230,10 @@ func ParseCopySourceRange(size int64, acceptRange string) (int64, int64, error)
 // ParseCopySource parses x-amz-copy-source header and returns source bucket,
 // source object, versionId, error respectively
 func ParseCopySource(copySourceHeader string) (string, string, string, error) {
+	if copySourceHeader == "" {
+		return "", "", "", s3err.GetInvalidArgumentErr(s3err.InvalidArgCopySourceBucket, copySourceHeader)
+	}
+
 	if copySourceHeader[0] == '/' {
 		copySourceHeader = copySourceHeader[1:]
 	}

backend/common_test.go

@@ -223,6 +223,15 @@ func TestParseCopySource(t *testing.T) {
 			wantErr:          true,
 			wantErrValue:     s3err.GetInvalidArgumentErr(s3err.InvalidArgCopySourceEncoding, "mybucket/object%ZZ"),
 		},
+		{
+			name:             "empty string",
+			copySourceHeader: "",
+			wantBucket:       "",
+			wantObject:       "",
+			wantVersionId:    "",
+			wantErr:          true,
+			wantErrValue:     s3err.GetInvalidArgumentErr(s3err.InvalidArgCopySourceBucket, ""),
+		},
 		{
 			name:             "missing object",
 			copySourceHeader: "mybucket",