Correct build-time directory creation with user namespaced daemon

This fixes errors in ownership on directory creation during build that
can cause inaccessible files depending on the paths in the Dockerfile
and non-existing directories in the starting image.

Add tests for the mkdir variants in pkg/idtools

Docker-DCO-1.1-Signed-off-by: Phil Estes <estesp@linux.vnet.ibm.com> (github: estesp)

idtools/usergroupadd_linux.go

@@ -2,13 +2,10 @@ package idtools
 
 import (
 	"fmt"
-	"os"
 	"os/exec"
 	"path/filepath"
 	"strings"
 	"syscall"
-
-	"github.com/docker/docker/pkg/system"
 )
 
 // add a user and/or group to Linux /etc/passwd, /etc/group using standard
@@ -156,20 +153,3 @@ func findUnused(file string, id int) (int, error) {
 		}
 	}
 }
-
-func mkdirAs(path string, mode os.FileMode, ownerUID, ownerGID int, mkAll bool) error {
-	if mkAll {
-		if err := system.MkdirAll(path, mode); err != nil && !os.IsExist(err) {
-			return err
-		}
-	} else {
-		if err := os.Mkdir(path, mode); err != nil && !os.IsExist(err) {
-			return err
-		}
-	}
-	// even if it existed, we will chown to change ownership as requested
-	if err := os.Chown(path, ownerUID, ownerGID); err != nil {
-		return err
-	}
-	return nil
-}