Skip UTF-8 BOM bytes from Dockerfile if exists

This fix tries to address issues in #23221 where Dockerfile
may consists of UTF-8 BOM. This likely happens when Notepad
tries to save a file as UTF-8 in Windows.

This fix skips the UTF-8 BOM bytes from the beginning of the
Dockerfile if exists.

Additional tests has been added to cover the changes in this
fix.

This fix fixes #23221.

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>

builder/dockerfile/parser/parser.go

@@ -3,6 +3,7 @@ package parser
 
 import (
 	"bufio"
+	"bytes"
 	"fmt"
 	"io"
 	"regexp"
@@ -152,8 +153,14 @@ func Parse(rwc io.Reader) (*Node, error) {
 	root.StartLine = -1
 	scanner := bufio.NewScanner(rwc)
 
+	utf8bom := []byte{0xEF, 0xBB, 0xBF}
 	for scanner.Scan() {
-		scannedLine := strings.TrimLeftFunc(scanner.Text(), unicode.IsSpace)
+		scannedBytes := scanner.Bytes()
+		// We trim UTF8 BOM
+		if currentLine == 0 {
+			scannedBytes = bytes.TrimPrefix(scannedBytes, utf8bom)
+		}
+		scannedLine := strings.TrimLeftFunc(string(scannedBytes), unicode.IsSpace)
 		currentLine++
 		line, child, err := ParseLine(scannedLine)
 		if err != nil {

integration-cli/docker_cli_build_test.go

@@ -6791,3 +6791,17 @@ foo2
 		c.Fatal(err)
 	}
 }
+
+// Test case for #23221
+func (s *DockerSuite) TestBuildWithUTF8BOM(c *check.C) {
+	name := "test-with-utf8-bom"
+	dockerfile := []byte(`FROM busybox`)
+	bomDockerfile := append([]byte{0xEF, 0xBB, 0xBF}, dockerfile...)
+	ctx, err := fakeContextFromNewTempDir()
+	c.Assert(err, check.IsNil)
+	defer ctx.Close()
+	err = ctx.addFile("Dockerfile", bomDockerfile)
+	c.Assert(err, check.IsNil)
+	_, err = buildImageFromContext(name, ctx, true)
+	c.Assert(err, check.IsNil)
+}