Merge pull request #228 from rafaeljusto/invalid-idna

check label for invalid code points

idn/punycode.go

@@ -3,9 +3,10 @@ package idn
 
 import (
 	"bytes"
-	"github.com/miekg/dns"
 	"strings"
 	"unicode"
+
+	"github.com/miekg/dns"
 )
 
 // Implementation idea from RFC itself and from from IDNA::Punycode created by
@@ -144,6 +145,10 @@ func encode(input []byte) []byte {
 
 	b := bytes.Runes(input)
 	for i := range b {
+		if !isValidRune(b[i]) {
+			return input
+		}
+
 		b[i] = preprune(b[i])
 	}
 
@@ -267,3 +272,34 @@ func decode(b []byte) []byte {
 	}
 	return ret.Bytes()
 }
+
+// isValidRune checks if the character is valid. We will look for the
+// character property in the code points list. For now we aren't checking special
+// rules in case of contextual property
+func isValidRune(r rune) bool {
+	return findProperty(r) == propertyPVALID
+}
+
+// findProperty will try to check the code point property of the given
+// character. It will use a binary search algorithm as we have a slice of
+// ordered ranges (average case performance O(log n))
+func findProperty(r rune) property {
+	imin, imax := 0, len(codePoints)
+
+	for imax >= imin {
+		imid := (imin + imax) / 2
+
+		codePoint := codePoints[imid]
+		if (codePoint.start == r && codePoint.end == 0) || (codePoint.start <= r && codePoint.end >= r) {
+			return codePoint.state
+		}
+
+		if (codePoint.end > 0 && codePoint.end < r) || (codePoint.end == 0 && codePoint.start < r) {
+			imin = imid + 1
+		} else {
+			imax = imid - 1
+		}
+	}
+
+	return propertyUnknown
+}

idn/punycode_test.go

@@ -13,13 +13,13 @@ var testcases = [][2]string{
 	{"AbC", "abc"},
 	{"я", "xn--41a"},
 	{"zя", "xn--z-0ub"},
-	{"ЯZ", "xn--z-zub"},
+	{"яZ", "xn--z-zub"},
 	{"а-я", "xn----7sb8g"},
 	{"إختبار", "xn--kgbechtv"},
 	{"آزمایشی", "xn--hgbk6aj7f53bba"},
 	{"测试", "xn--0zwm56d"},
 	{"測試", "xn--g6w251d"},
-	{"Испытание", "xn--80akhbyknj4f"},
+	{"испытание", "xn--80akhbyknj4f"},
 	{"परीक्षा", "xn--11b5bs3a9aj6g"},
 	{"δοκιμή", "xn--jxalpdlp"},
 	{"테스트", "xn--9t4b11yi5a"},
@@ -27,6 +27,7 @@ var testcases = [][2]string{
 	{"テスト", "xn--zckzah"},
 	{"பரிட்சை", "xn--hlcj6aya9esc7a"},
 	{"mamão-com-açúcar", "xn--mamo-com-acar-yeb1e6q"},
+	{"σ", "xn--4xa"},
 }
 
 func TestEncodeDecodePunycode(t *testing.T) {
@@ -81,17 +82,34 @@ func TestEncodeDecodeFinalPeriod(t *testing.T) {
 	}
 }
 
-var invalid = []string{
+var invalidACEs = []string{
 	"xn--*",
 	"xn--",
 	"xn---",
 }
 
 func TestInvalidPunycode(t *testing.T) {
-	for _, d := range invalid {
+	for _, d := range invalidACEs {
 		s := FromPunycode(d)
 		if s != d {
 			t.Errorf("Changed invalid name %s to %#v", d, s)
 		}
 	}
 }
+
+// You can verify the labels that are valid or not comparing to the Verisign
+// website: http://mct.verisign-grs.com/
+var invalidUnicodes = []string{
+	"Σ",
+	"ЯZ",
+	"Испытание",
+}
+
+func TestInvalidUnicodes(t *testing.T) {
+	for _, d := range invalidUnicodes {
+		s := ToPunycode(d)
+		if s != d {
+			t.Errorf("Changed invalid name %s to %#v", d, s)
+		}
+	}
+}