XL: explicit deleteObject is not needed after rename failure. (#1760)

Reason is renameObject() does deleteObject() upon writeQuorum
failure if not keeps the successfully renamed parts if we have
reached readQuorum.

xl-v1-multipart.go

@@ -88,9 +88,6 @@ func (xl xlObjects) newMultipartUploadCommon(bucket string, object string, meta
 	}
 	rErr := xl.renameObject(minioMetaBucket, tempUploadIDPath, minioMetaBucket, uploadIDPath)
 	if rErr == nil {
-		if dErr := xl.deleteObject(minioMetaBucket, tempUploadIDPath); dErr != nil {
-			return "", toObjectErr(dErr, minioMetaBucket, tempUploadIDPath)
-		}
 		// Return success.
 		return uploadID, nil
 	}
@@ -194,9 +191,6 @@ func (xl xlObjects) putObjectPartCommon(bucket string, object string, uploadID s
 	partPath := path.Join(mpartMetaPrefix, bucket, object, uploadID, partSuffix)
 	err = xl.renameObject(minioMetaBucket, tmpPartPath, minioMetaBucket, partPath)
 	if err != nil {
-		if dErr := xl.deleteObject(minioMetaBucket, tmpPartPath); dErr != nil {
-			return "", toObjectErr(dErr, minioMetaBucket, tmpPartPath)
-		}
 		return "", toObjectErr(err, minioMetaBucket, partPath)
 	}
 	if err = xl.writeXLMetadata(minioMetaBucket, path.Join(mpartMetaPrefix, bucket, object, uploadID), xlMeta); err != nil {

xl-v1-object.go

@@ -269,9 +269,6 @@ func (xl xlObjects) PutObject(bucket string, object string, size int64, data io.
 
 	err = xl.renameObject(minioMetaBucket, tempObj, bucket, object)
 	if err != nil {
-		if dErr := xl.deleteObject(minioMetaBucket, tempObj); dErr != nil {
-			return "", toObjectErr(dErr, minioMetaBucket, tempObj)
-		}
 		return "", toObjectErr(err, bucket, object)
 	}
 
@@ -300,7 +297,12 @@ func (xl xlObjects) deleteObject(bucket, object string) error {
 		wg.Add(1)
 		go func(index int, disk StorageAPI) {
 			defer wg.Done()
-			dErrs[index] = cleanupDir(disk, bucket, object)
+			err := cleanupDir(disk, bucket, object)
+			if err != nil {
+				dErrs[index] = err
+				return
+			}
+			dErrs[index] = nil
 		}(index, disk)
 	}