fix: stop EscapeDN and PostalAddress.Escape corrupting invalid UTF-8

v3/ldap.go

@@ -1,10 +1,12 @@
 package ldap
 
 import (
+	"encoding/hex"
 	"fmt"
 	"log"
 	"os"
 	"strings"
+	"unicode/utf8"
 
 	ber "github.com/go-asn1-ber/asn1-ber"
 )
@@ -489,6 +491,18 @@ func EscapeDN(dn string) string {
 	builder := strings.Builder{}
 
 	for i, r := range dn {
+		// A byte that is not part of a valid UTF-8 sequence ranges as
+		// (utf8.RuneError, width 1). Writing the rune would silently replace
+		// the byte with U+FFFD, so two different values could escape to the
+		// same DN; escape the raw byte as an RFC 4514 hex pair instead.
+		if r == utf8.RuneError {
+			if _, size := utf8.DecodeRuneInString(dn[i:]); size == 1 {
+				builder.WriteByte('\\')
+				builder.WriteString(hex.EncodeToString([]byte{dn[i]}))
+				continue
+			}
+		}
+
 		// Escape leading and trailing spaces
 		if (i == 0 || i == len(dn)-1) && r == ' ' {
 			builder.WriteRune('\\')

v3/ldap_test.go

@@ -348,6 +348,9 @@ func TestEscapeDN(t *testing.T) {
 		{name: "nullByte", dn: "\u0000te\x00st\x00user" + string(rune(0)), want: "\\00te\\00st\\00user\\00"},
 		{name: "variousCharacters", dn: "test\"+,;<>\\-_user", want: "test\\\"\\+\\,\\;\\<\\>\\\\-_user"},
 		{name: "multiByteRunes", dn: "test\u0391user ", want: "test\u0391user\\ "},
+		{name: "invalidUTF8Byte", dn: "test\xebuser", want: "test\\ebuser"},
+		{name: "truncatedUTF8Sequence", dn: "test\xc3", want: "test\\c3"},
+		{name: "replacementCharKeptDistinct", dn: "test\ufffduser", want: "test\ufffduser"},
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {

v3/postaladdress.go

@@ -49,14 +49,19 @@ func (p *PostalAddress) Escape() string {
 	builder := &strings.Builder{}
 
 	for _, line := range p.lines {
-		for _, char := range line {
-			switch char {
+		// Iterate bytes, not runes: ranging a line that is not valid UTF-8
+		// would replace each invalid byte with U+FFFD, so two different lines
+		// could escape to the same value. The escaped characters are ASCII and
+		// ParsePostalAddress reads bytes, so every other octet passes through
+		// unchanged.
+		for i := 0; i < len(line); i++ {
+			switch line[i] {
 			case '\\':
 				builder.WriteString("\\5C")
 			case '$':
 				builder.WriteString("\\24")
 			default:
-				builder.WriteRune(char)
+				builder.WriteByte(line[i])
 			}
 		}
 

v3/postaladdress_test.go

@@ -200,6 +200,19 @@ func TestPostalAddressEqual(t *testing.T) {
 	}
 }
 
+func TestPostalAddressEscapeInvalidUTF8(t *testing.T) {
+	// A line that is not valid UTF-8 must keep its original bytes when
+	// escaped; replacing them with U+FFFD would make two different lines
+	// escape to the same value.
+	addr := mustNewPostalAddress(t, []string{"caf\xe9 $1"})
+	escaped := addr.Escape()
+	assert.Equal(t, "caf\xe9 \\241$", escaped)
+
+	parsed, err := ParsePostalAddress(escaped)
+	assert.NoError(t, err)
+	assert.True(t, addr.Equal(parsed), "escaped form should parse back to the same lines")
+}
+
 func TestParsePostalAddress_Escape(t *testing.T) {
 	t.Run("incomplete escape", func(t *testing.T) {
 		_, err := ParsePostalAddress("AAAAAAAAAA\\")