Switch to using the dep tool and update all the dependencies

vendor/golang.org/x/text/unicode/norm/composition.go

@@ -33,17 +33,9 @@ const (
 // streamSafe implements the policy of when a CGJ should be inserted.
 type streamSafe uint8
 
-// mkStreamSafe is a shorthand for declaring a streamSafe var and calling
-// first on it.
-func mkStreamSafe(p Properties) streamSafe {
-	return streamSafe(p.nTrailingNonStarters())
-}
-
-// first inserts the first rune of a segment.
+// first inserts the first rune of a segment. It is a faster version of next if
+// it is known p represents the first rune in a segment.
 func (ss *streamSafe) first(p Properties) {
-	if *ss != 0 {
-		panic("!= 0")
-	}
 	*ss = streamSafe(p.nTrailingNonStarters())
 }
 
@@ -66,7 +58,7 @@ func (ss *streamSafe) next(p Properties) ssState {
 	// be a non-starter. Note that it always hold that if nLead > 0 then
 	// nLead == nTrail.
 	if n == 0 {
-		*ss = 0
+		*ss = streamSafe(p.nTrailingNonStarters())
 		return ssStarter
 	}
 	return ssSuccess
@@ -142,7 +134,6 @@ func (rb *reorderBuffer) setFlusher(out []byte, f func(*reorderBuffer) bool) {
 func (rb *reorderBuffer) reset() {
 	rb.nrune = 0
 	rb.nbyte = 0
-	rb.ss = 0
 }
 
 func (rb *reorderBuffer) doFlush() bool {
@@ -257,6 +248,9 @@ func (rb *reorderBuffer) insertUnsafe(src input, i int, info Properties) {
 // It flushes the buffer on each new segment start.
 func (rb *reorderBuffer) insertDecomposed(dcomp []byte) insertErr {
 	rb.tmpBytes.setBytes(dcomp)
+	// As the streamSafe accounting already handles the counting for modifiers,
+	// we don't have to call next. However, we do need to keep the accounting
+	// intact when flushing the buffer.
 	for i := 0; i < len(dcomp); {
 		info := rb.f.info(rb.tmpBytes, i)
 		if info.BoundaryBefore() && rb.nrune > 0 && !rb.doFlush() {

vendor/golang.org/x/text/unicode/norm/composition_test.go

@@ -0,0 +1,130 @@
+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package norm
+
+import "testing"
+
+// TestCase is used for most tests.
+type TestCase struct {
+	in  []rune
+	out []rune
+}
+
+func runTests(t *testing.T, name string, fm Form, tests []TestCase) {
+	rb := reorderBuffer{}
+	rb.init(fm, nil)
+	for i, test := range tests {
+		rb.setFlusher(nil, appendFlush)
+		for j, rune := range test.in {
+			b := []byte(string(rune))
+			src := inputBytes(b)
+			info := rb.f.info(src, 0)
+			if j == 0 {
+				rb.ss.first(info)
+			} else {
+				rb.ss.next(info)
+			}
+			if rb.insertFlush(src, 0, info) < 0 {
+				t.Errorf("%s:%d: insert failed for rune %d", name, i, j)
+			}
+		}
+		rb.doFlush()
+		was := string(rb.out)
+		want := string(test.out)
+		if len(was) != len(want) {
+			t.Errorf("%s:%d: length = %d; want %d", name, i, len(was), len(want))
+		}
+		if was != want {
+			k, pfx := pidx(was, want)
+			t.Errorf("%s:%d: \nwas  %s%+q; \nwant %s%+q", name, i, pfx, was[k:], pfx, want[k:])
+		}
+	}
+}
+
+func TestFlush(t *testing.T) {
+	const (
+		hello = "Hello "
+		world = "world!"
+	)
+	buf := make([]byte, maxByteBufferSize)
+	p := copy(buf, hello)
+	out := buf[p:]
+	rb := reorderBuffer{}
+	rb.initString(NFC, world)
+	if i := rb.flushCopy(out); i != 0 {
+		t.Errorf("wrote bytes on flush of empty buffer. (len(out) = %d)", i)
+	}
+
+	for i := range world {
+		// No need to set streamSafe values for this test.
+		rb.insertFlush(rb.src, i, rb.f.info(rb.src, i))
+		n := rb.flushCopy(out)
+		out = out[n:]
+		p += n
+	}
+
+	was := buf[:p]
+	want := hello + world
+	if string(was) != want {
+		t.Errorf(`output after flush was "%s"; want "%s"`, string(was), want)
+	}
+	if rb.nrune != 0 {
+		t.Errorf("non-null size of info buffer (rb.nrune == %d)", rb.nrune)
+	}
+	if rb.nbyte != 0 {
+		t.Errorf("non-null size of byte buffer (rb.nbyte == %d)", rb.nbyte)
+	}
+}
+
+var insertTests = []TestCase{
+	{[]rune{'a'}, []rune{'a'}},
+	{[]rune{0x300}, []rune{0x300}},
+	{[]rune{0x300, 0x316}, []rune{0x316, 0x300}}, // CCC(0x300)==230; CCC(0x316)==220
+	{[]rune{0x316, 0x300}, []rune{0x316, 0x300}},
+	{[]rune{0x41, 0x316, 0x300}, []rune{0x41, 0x316, 0x300}},
+	{[]rune{0x41, 0x300, 0x316}, []rune{0x41, 0x316, 0x300}},
+	{[]rune{0x300, 0x316, 0x41}, []rune{0x316, 0x300, 0x41}},
+	{[]rune{0x41, 0x300, 0x40, 0x316}, []rune{0x41, 0x300, 0x40, 0x316}},
+}
+
+func TestInsert(t *testing.T) {
+	runTests(t, "TestInsert", NFD, insertTests)
+}
+
+var decompositionNFDTest = []TestCase{
+	{[]rune{0xC0}, []rune{0x41, 0x300}},
+	{[]rune{0xAC00}, []rune{0x1100, 0x1161}},
+	{[]rune{0x01C4}, []rune{0x01C4}},
+	{[]rune{0x320E}, []rune{0x320E}},
+	{[]rune("음ẻ과"), []rune{0x110B, 0x1173, 0x11B7, 0x65, 0x309, 0x1100, 0x116A}},
+}
+
+var decompositionNFKDTest = []TestCase{
+	{[]rune{0xC0}, []rune{0x41, 0x300}},
+	{[]rune{0xAC00}, []rune{0x1100, 0x1161}},
+	{[]rune{0x01C4}, []rune{0x44, 0x5A, 0x030C}},
+	{[]rune{0x320E}, []rune{0x28, 0x1100, 0x1161, 0x29}},
+}
+
+func TestDecomposition(t *testing.T) {
+	runTests(t, "TestDecompositionNFD", NFD, decompositionNFDTest)
+	runTests(t, "TestDecompositionNFKD", NFKD, decompositionNFKDTest)
+}
+
+var compositionTest = []TestCase{
+	{[]rune{0x41, 0x300}, []rune{0xC0}},
+	{[]rune{0x41, 0x316}, []rune{0x41, 0x316}},
+	{[]rune{0x41, 0x300, 0x35D}, []rune{0xC0, 0x35D}},
+	{[]rune{0x41, 0x316, 0x300}, []rune{0xC0, 0x316}},
+	// blocking starter
+	{[]rune{0x41, 0x316, 0x40, 0x300}, []rune{0x41, 0x316, 0x40, 0x300}},
+	{[]rune{0x1100, 0x1161}, []rune{0xAC00}},
+	// parenthesized Hangul, alternate between ASCII and Hangul.
+	{[]rune{0x28, 0x1100, 0x1161, 0x29}, []rune{0x28, 0xAC00, 0x29}},
+}
+
+func TestComposition(t *testing.T) {
+	runTests(t, "TestComposition", NFC, compositionTest)
+}

vendor/golang.org/x/text/unicode/norm/example_iter_test.go

@@ -0,0 +1,82 @@
+// Copyright 2012 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package norm_test
+
+import (
+	"bytes"
+	"fmt"
+	"unicode/utf8"
+
+	"golang.org/x/text/unicode/norm"
+)
+
+// EqualSimple uses a norm.Iter to compare two non-normalized
+// strings for equivalence.
+func EqualSimple(a, b string) bool {
+	var ia, ib norm.Iter
+	ia.InitString(norm.NFKD, a)
+	ib.InitString(norm.NFKD, b)
+	for !ia.Done() && !ib.Done() {
+		if !bytes.Equal(ia.Next(), ib.Next()) {
+			return false
+		}
+	}
+	return ia.Done() && ib.Done()
+}
+
+// FindPrefix finds the longest common prefix of ASCII characters
+// of a and b.
+func FindPrefix(a, b string) int {
+	i := 0
+	for ; i < len(a) && i < len(b) && a[i] < utf8.RuneSelf && a[i] == b[i]; i++ {
+	}
+	return i
+}
+
+// EqualOpt is like EqualSimple, but optimizes the special
+// case for ASCII characters.
+func EqualOpt(a, b string) bool {
+	n := FindPrefix(a, b)
+	a, b = a[n:], b[n:]
+	var ia, ib norm.Iter
+	ia.InitString(norm.NFKD, a)
+	ib.InitString(norm.NFKD, b)
+	for !ia.Done() && !ib.Done() {
+		if !bytes.Equal(ia.Next(), ib.Next()) {
+			return false
+		}
+		if n := int64(FindPrefix(a[ia.Pos():], b[ib.Pos():])); n != 0 {
+			ia.Seek(n, 1)
+			ib.Seek(n, 1)
+		}
+	}
+	return ia.Done() && ib.Done()
+}
+
+var compareTests = []struct{ a, b string }{
+	{"aaa", "aaa"},
+	{"aaa", "aab"},
+	{"a\u0300a", "\u00E0a"},
+	{"a\u0300\u0320b", "a\u0320\u0300b"},
+	{"\u1E0A\u0323", "\x44\u0323\u0307"},
+	// A character that decomposes into multiple segments
+	// spans several iterations.
+	{"\u3304", "\u30A4\u30CB\u30F3\u30AF\u3099"},
+}
+
+func ExampleIter() {
+	for i, t := range compareTests {
+		r0 := EqualSimple(t.a, t.b)
+		r1 := EqualOpt(t.a, t.b)
+		fmt.Printf("%d: %v %v\n", i, r0, r1)
+	}
+	// Output:
+	// 0: true true
+	// 1: false false
+	// 2: true true
+	// 3: true true
+	// 4: true true
+	// 5: true true
+}

vendor/golang.org/x/text/unicode/norm/example_test.go

@@ -0,0 +1,27 @@
+// Copyright 2016 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package norm_test
+
+import (
+	"fmt"
+
+	"golang.org/x/text/unicode/norm"
+)
+
+func ExampleForm_NextBoundary() {
+	s := norm.NFD.String("Mêlée")
+
+	for i := 0; i < len(s); {
+		d := norm.NFC.NextBoundaryInString(s[i:], true)
+		fmt.Printf("%[1]s: %+[1]q\n", s[i:i+d])
+		i += d
+	}
+	// Output:
+	// M: "M"
+	// ê: "e\u0302"
+	// l: "l"
+	// é: "e\u0301"
+	// e: "e"
+}

vendor/golang.org/x/text/unicode/norm/forminfo.go

@@ -10,7 +10,7 @@ package norm
 // and its corresponding decomposing form share the same trie.  Each trie maps
 // a rune to a uint16. The values take two forms.  For v >= 0x8000:
 //   bits
-//   15:    1 (inverse of NFD_QD bit of qcInfo)
+//   15:    1 (inverse of NFD_QC bit of qcInfo)
 //   13..7: qcInfo (see below). isYesD is always true (no decompostion).
 //    6..0: ccc (compressed CCC value).
 // For v < 0x8000, the respective rune has a decomposition and v is an index

vendor/golang.org/x/text/unicode/norm/forminfo_test.go

@@ -0,0 +1,54 @@
+// Copyright 2013 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// +build test
+
+package norm
+
+import "testing"
+
+func TestProperties(t *testing.T) {
+	var d runeData
+	CK := [2]string{"C", "K"}
+	for k, r := 1, rune(0); r < 0x2ffff; r++ {
+		if k < len(testData) && r == testData[k].r {
+			d = testData[k]
+			k++
+		}
+		s := string(r)
+		for j, p := range []Properties{NFC.PropertiesString(s), NFKC.PropertiesString(s)} {
+			f := d.f[j]
+			if p.CCC() != d.ccc {
+				t.Errorf("%U: ccc(%s): was %d; want %d %X", r, CK[j], p.CCC(), d.ccc, p.index)
+			}
+			if p.isYesC() != (f.qc == Yes) {
+				t.Errorf("%U: YesC(%s): was %v; want %v", r, CK[j], p.isYesC(), f.qc == Yes)
+			}
+			if p.combinesBackward() != (f.qc == Maybe) {
+				t.Errorf("%U: combines backwards(%s): was %v; want %v", r, CK[j], p.combinesBackward(), f.qc == Maybe)
+			}
+			if p.nLeadingNonStarters() != d.nLead {
+				t.Errorf("%U: nLead(%s): was %d; want %d %#v %#v", r, CK[j], p.nLeadingNonStarters(), d.nLead, p, d)
+			}
+			if p.nTrailingNonStarters() != d.nTrail {
+				t.Errorf("%U: nTrail(%s): was %d; want %d %#v %#v", r, CK[j], p.nTrailingNonStarters(), d.nTrail, p, d)
+			}
+			if p.combinesForward() != f.combinesForward {
+				t.Errorf("%U: combines forward(%s): was %v; want %v %#v", r, CK[j], p.combinesForward(), f.combinesForward, p)
+			}
+			// Skip Hangul as it is algorithmically computed.
+			if r >= hangulBase && r < hangulEnd {
+				continue
+			}
+			if p.hasDecomposition() {
+				if has := f.decomposition != ""; !has {
+					t.Errorf("%U: hasDecomposition(%s): was %v; want %v", r, CK[j], p.hasDecomposition(), has)
+				}
+				if string(p.Decomposition()) != f.decomposition {
+					t.Errorf("%U: decomp(%s): was %+q; want %+q", r, CK[j], p.Decomposition(), f.decomposition)
+				}
+			}
+		}
+	}
+}

vendor/golang.org/x/text/unicode/norm/input.go

@@ -90,16 +90,20 @@ func (in *input) charinfoNFKC(p int) (uint16, int) {
 }
 
 func (in *input) hangul(p int) (r rune) {
+	var size int
 	if in.bytes == nil {
 		if !isHangulString(in.str[p:]) {
 			return 0
 		}
-		r, _ = utf8.DecodeRuneInString(in.str[p:])
+		r, size = utf8.DecodeRuneInString(in.str[p:])
 	} else {
 		if !isHangul(in.bytes[p:]) {
 			return 0
 		}
-		r, _ = utf8.DecodeRune(in.bytes[p:])
+		r, size = utf8.DecodeRune(in.bytes[p:])
+	}
+	if size != hangulUTF8Size {
+		return 0
 	}
 	return r
 }

vendor/golang.org/x/text/unicode/norm/iter.go

@@ -41,6 +41,7 @@ func (i *Iter) Init(f Form, src []byte) {
 	i.next = i.rb.f.nextMain
 	i.asciiF = nextASCIIBytes
 	i.info = i.rb.f.info(i.rb.src, i.p)
+	i.rb.ss.first(i.info)
 }
 
 // InitString initializes i to iterate over src after normalizing it to Form f.
@@ -56,11 +57,12 @@ func (i *Iter) InitString(f Form, src string) {
 	i.next = i.rb.f.nextMain
 	i.asciiF = nextASCIIString
 	i.info = i.rb.f.info(i.rb.src, i.p)
+	i.rb.ss.first(i.info)
 }
 
 // Seek sets the segment to be returned by the next call to Next to start
 // at position p.  It is the responsibility of the caller to set p to the
-// start of a UTF8 rune.
+// start of a segment.
 func (i *Iter) Seek(offset int64, whence int) (int64, error) {
 	var abs int64
 	switch whence {
@@ -84,6 +86,7 @@ func (i *Iter) Seek(offset int64, whence int) (int64, error) {
 	i.multiSeg = nil
 	i.next = i.rb.f.nextMain
 	i.info = i.rb.f.info(i.rb.src, i.p)
+	i.rb.ss.first(i.info)
 	return abs, nil
 }
 
@@ -161,6 +164,7 @@ func nextHangul(i *Iter) []byte {
 	if next >= i.rb.nsrc {
 		i.setDone()
 	} else if i.rb.src.hangul(next) == 0 {
+		i.rb.ss.next(i.info)
 		i.info = i.rb.f.info(i.rb.src, i.p)
 		i.next = i.rb.f.nextMain
 		return i.next(i)
@@ -204,12 +208,10 @@ func nextMultiNorm(i *Iter) []byte {
 		if info.BoundaryBefore() {
 			i.rb.compose()
 			seg := i.buf[:i.rb.flushCopy(i.buf[:])]
-			i.rb.ss.first(info)
 			i.rb.insertUnsafe(input{bytes: d}, j, info)
 			i.multiSeg = d[j+int(info.size):]
 			return seg
 		}
-		i.rb.ss.next(info)
 		i.rb.insertUnsafe(input{bytes: d}, j, info)
 		j += int(info.size)
 	}
@@ -222,9 +224,9 @@ func nextMultiNorm(i *Iter) []byte {
 func nextDecomposed(i *Iter) (next []byte) {
 	outp := 0
 	inCopyStart, outCopyStart := i.p, 0
-	ss := mkStreamSafe(i.info)
 	for {
 		if sz := int(i.info.size); sz <= 1 {
+			i.rb.ss = 0
 			p := i.p
 			i.p++ // ASCII or illegal byte.  Either way, advance by 1.
 			if i.p >= i.rb.nsrc {
@@ -243,6 +245,8 @@ func nextDecomposed(i *Iter) (next []byte) {
 			p := outp + len(d)
 			if outp > 0 {
 				i.rb.src.copySlice(i.buf[outCopyStart:], inCopyStart, i.p)
+				// TODO: this condition should not be possible, but we leave it
+				// in for defensive purposes.
 				if p > len(i.buf) {
 					return i.buf[:outp]
 				}
@@ -266,7 +270,7 @@ func nextDecomposed(i *Iter) (next []byte) {
 			} else {
 				i.info = i.rb.f.info(i.rb.src, i.p)
 			}
-			switch ss.next(i.info) {
+			switch i.rb.ss.next(i.info) {
 			case ssOverflow:
 				i.next = nextCGJDecompose
 				fallthrough
@@ -309,7 +313,7 @@ func nextDecomposed(i *Iter) (next []byte) {
 		}
 		prevCC := i.info.tccc
 		i.info = i.rb.f.info(i.rb.src, i.p)
-		if v := ss.next(i.info); v == ssStarter {
+		if v := i.rb.ss.next(i.info); v == ssStarter {
 			break
 		} else if v == ssOverflow {
 			i.next = nextCGJDecompose
@@ -335,10 +339,6 @@ doNorm:
 
 func doNormDecomposed(i *Iter) []byte {
 	for {
-		if s := i.rb.ss.next(i.info); s == ssOverflow {
-			i.next = nextCGJDecompose
-			break
-		}
 		i.rb.insertUnsafe(i.rb.src, i.p, i.info)
 		if i.p += int(i.info.size); i.p >= i.rb.nsrc {
 			i.setDone()
@@ -348,6 +348,10 @@ func doNormDecomposed(i *Iter) []byte {
 		if i.info.ccc == 0 {
 			break
 		}
+		if s := i.rb.ss.next(i.info); s == ssOverflow {
+			i.next = nextCGJDecompose
+			break
+		}
 	}
 	// new segment or too many combining characters: exit normalization
 	return i.buf[:i.rb.flushCopy(i.buf[:])]
@@ -357,6 +361,7 @@ func nextCGJDecompose(i *Iter) []byte {
 	i.rb.ss = 0
 	i.rb.insertCGJ()
 	i.next = nextDecomposed
+	i.rb.ss.first(i.info)
 	buf := doNormDecomposed(i)
 	return buf
 }
@@ -365,7 +370,6 @@ func nextCGJDecompose(i *Iter) []byte {
 func nextComposed(i *Iter) []byte {
 	outp, startp := 0, i.p
 	var prevCC uint8
-	ss := mkStreamSafe(i.info)
 	for {
 		if !i.info.isYesC() {
 			goto doNorm
@@ -385,11 +389,12 @@ func nextComposed(i *Iter) []byte {
 			i.setDone()
 			break
 		} else if i.rb.src._byte(i.p) < utf8.RuneSelf {
+			i.rb.ss = 0
 			i.next = i.asciiF
 			break
 		}
 		i.info = i.rb.f.info(i.rb.src, i.p)
-		if v := ss.next(i.info); v == ssStarter {
+		if v := i.rb.ss.next(i.info); v == ssStarter {
 			break
 		} else if v == ssOverflow {
 			i.next = nextCGJCompose
@@ -401,8 +406,10 @@ func nextComposed(i *Iter) []byte {
 	}
 	return i.returnSlice(startp, i.p)
 doNorm:
+	// reset to start position
 	i.p = startp
 	i.info = i.rb.f.info(i.rb.src, i.p)
+	i.rb.ss.first(i.info)
 	if i.info.multiSegment() {
 		d := i.info.Decomposition()
 		info := i.rb.f.info(input{bytes: d}, 0)

vendor/golang.org/x/text/unicode/norm/iter_test.go

@@ -0,0 +1,98 @@
+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package norm
+
+import (
+	"strings"
+	"testing"
+)
+
+func doIterNorm(f Form, s string) []byte {
+	acc := []byte{}
+	i := Iter{}
+	i.InitString(f, s)
+	for !i.Done() {
+		acc = append(acc, i.Next()...)
+	}
+	return acc
+}
+
+func TestIterNext(t *testing.T) {
+	runNormTests(t, "IterNext", func(f Form, out []byte, s string) []byte {
+		return doIterNorm(f, string(append(out, s...)))
+	})
+}
+
+type SegmentTest struct {
+	in  string
+	out []string
+}
+
+var segmentTests = []SegmentTest{
+	{"\u1E0A\u0323a", []string{"\x44\u0323\u0307", "a", ""}},
+	{rep('a', segSize), append(strings.Split(rep('a', segSize), ""), "")},
+	{rep('a', segSize+2), append(strings.Split(rep('a', segSize+2), ""), "")},
+	{rep('a', segSize) + "\u0300aa",
+		append(strings.Split(rep('a', segSize-1), ""), "a\u0300", "a", "a", "")},
+
+	// U+0f73 is NOT treated as a starter as it is a modifier
+	{"a" + grave(29) + "\u0f73", []string{"a" + grave(29), cgj + "\u0f73"}},
+	{"a\u0f73", []string{"a\u0f73"}},
+
+	// U+ff9e is treated as a non-starter.
+	// TODO: should we? Note that this will only affect iteration, as whether
+	// or not we do so does not affect the normalization output and will either
+	// way result in consistent iteration output.
+	{"a" + grave(30) + "\uff9e", []string{"a" + grave(30), cgj + "\uff9e"}},
+	{"a\uff9e", []string{"a\uff9e"}},
+}
+
+var segmentTestsK = []SegmentTest{
+	{"\u3332", []string{"\u30D5", "\u30A1", "\u30E9", "\u30C3", "\u30C8\u3099", ""}},
+	// last segment of multi-segment decomposition needs normalization
+	{"\u3332\u093C", []string{"\u30D5", "\u30A1", "\u30E9", "\u30C3", "\u30C8\u093C\u3099", ""}},
+	{"\u320E", []string{"\x28", "\uAC00", "\x29"}},
+
+	// last segment should be copied to start of buffer.
+	{"\ufdfa", []string{"\u0635", "\u0644", "\u0649", " ", "\u0627", "\u0644", "\u0644", "\u0647", " ", "\u0639", "\u0644", "\u064a", "\u0647", " ", "\u0648", "\u0633", "\u0644", "\u0645", ""}},
+	{"\ufdfa" + grave(30), []string{"\u0635", "\u0644", "\u0649", " ", "\u0627", "\u0644", "\u0644", "\u0647", " ", "\u0639", "\u0644", "\u064a", "\u0647", " ", "\u0648", "\u0633", "\u0644", "\u0645" + grave(30), ""}},
+	{"\uFDFA" + grave(64), []string{"\u0635", "\u0644", "\u0649", " ", "\u0627", "\u0644", "\u0644", "\u0647", " ", "\u0639", "\u0644", "\u064a", "\u0647", " ", "\u0648", "\u0633", "\u0644", "\u0645" + grave(30), cgj + grave(30), cgj + grave(4), ""}},
+
+	// Hangul and Jamo are grouped together.
+	{"\uAC00", []string{"\u1100\u1161", ""}},
+	{"\uAC01", []string{"\u1100\u1161\u11A8", ""}},
+	{"\u1100\u1161", []string{"\u1100\u1161", ""}},
+}
+
+// Note that, by design, segmentation is equal for composing and decomposing forms.
+func TestIterSegmentation(t *testing.T) {
+	segmentTest(t, "SegmentTestD", NFD, segmentTests)
+	segmentTest(t, "SegmentTestC", NFC, segmentTests)
+	segmentTest(t, "SegmentTestKD", NFKD, segmentTestsK)
+	segmentTest(t, "SegmentTestKC", NFKC, segmentTestsK)
+}
+
+func segmentTest(t *testing.T, name string, f Form, tests []SegmentTest) {
+	iter := Iter{}
+	for i, tt := range tests {
+		iter.InitString(f, tt.in)
+		for j, seg := range tt.out {
+			if seg == "" {
+				if !iter.Done() {
+					res := string(iter.Next())
+					t.Errorf(`%s:%d:%d: expected Done()==true, found segment %+q`, name, i, j, res)
+				}
+				continue
+			}
+			if iter.Done() {
+				t.Errorf("%s:%d:%d: Done()==true, want false", name, i, j)
+			}
+			seg = f.String(seg)
+			if res := string(iter.Next()); res != seg {
+				t.Errorf(`%s:%d:%d" segment was %+q (%d); want %+q (%d)`, name, i, j, pc(res), len(res), pc(seg), len(seg))
+			}
+		}
+	}
+}

vendor/golang.org/x/text/unicode/norm/maketables.go

@@ -35,12 +35,9 @@ func main() {
 	computeNonStarterCounts()
 	verifyComputed()
 	printChars()
-	if *test {
-		testDerived()
-		printTestdata()
-	} else {
-		makeTables()
-	}
+	testDerived()
+	printTestdata()
+	makeTables()
 }
 
 var (
@@ -602,6 +599,7 @@ func printCharInfoTables(w io.Writer) int {
 		}
 		index := normalDecomp
 		nTrail := chars[r].nTrailingNonStarters
+		nLead := chars[r].nLeadingNonStarters
 		if tccc > 0 || lccc > 0 || nTrail > 0 {
 			tccc <<= 2
 			tccc |= nTrail
@@ -612,7 +610,7 @@ func printCharInfoTables(w io.Writer) int {
 					index = firstCCC
 				}
 			}
-			if lccc > 0 {
+			if lccc > 0 || nLead > 0 {
 				s += string([]byte{lccc})
 				if index == firstCCC {
 					log.Fatalf("%U: multi-segment decomposition not supported for decompositions with leading CCC != 0", r)

vendor/golang.org/x/text/unicode/norm/norm_test.go

@@ -0,0 +1,14 @@
+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package norm_test
+
+import (
+	"testing"
+)
+
+func TestPlaceHolder(t *testing.T) {
+	// Does nothing, just allows the Makefile to be canonical
+	// while waiting for the package itself to be written.
+}

vendor/golang.org/x/text/unicode/norm/normalize.go

@@ -2,11 +2,12 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+// Note: the file data_test.go that is generated should not be checked in.
 //go:generate go run maketables.go triegen.go
-//go:generate go run maketables.go triegen.go -test
+//go:generate go test -tags test
 
 // Package norm contains types and functions for normalizing Unicode strings.
-package norm
+package norm // import "golang.org/x/text/unicode/norm"
 
 import (
 	"unicode/utf8"
@@ -323,7 +324,6 @@ func (f *formInfo) quickSpan(src input, i, end int, atEOF bool) (n int, ok bool)
 		// have an overflow for runes that are starters (e.g. with U+FF9E).
 		switch ss.next(info) {
 		case ssStarter:
-			ss.first(info)
 			lastSegStart = i
 		case ssOverflow:
 			return lastSegStart, false
@@ -440,6 +440,8 @@ func (f Form) nextBoundary(src input, nsrc int, atEOF bool) int {
 			}
 			return -1
 		}
+		// TODO: Using streamSafe to determine the boundary isn't the same as
+		// using BoundaryBefore. Determine which should be used.
 		if s := ss.next(info); s != ssSuccess {
 			return i
 		}
@@ -504,15 +506,14 @@ func decomposeSegment(rb *reorderBuffer, sp int, atEOF bool) int {
 	if info.size == 0 {
 		return 0
 	}
-	if rb.nrune > 0 {
-		if s := rb.ss.next(info); s == ssStarter {
-			goto end
-		} else if s == ssOverflow {
-			rb.insertCGJ()
+	if s := rb.ss.next(info); s == ssStarter {
+		// TODO: this could be removed if we don't support merging.
+		if rb.nrune > 0 {
 			goto end
 		}
-	} else {
-		rb.ss.first(info)
+	} else if s == ssOverflow {
+		rb.insertCGJ()
+		goto end
 	}
 	if err := rb.insertFlush(rb.src, sp, info); err != iSuccess {
 		return int(err)

vendor/golang.org/x/text/unicode/norm/readwriter_test.go

@@ -0,0 +1,56 @@
+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package norm
+
+import (
+	"bytes"
+	"fmt"
+	"testing"
+)
+
+var bufSizes = []int{1, 2, 3, 4, 5, 6, 7, 8, 100, 101, 102, 103, 4000, 4001, 4002, 4003}
+
+func readFunc(size int) appendFunc {
+	return func(f Form, out []byte, s string) []byte {
+		out = append(out, s...)
+		r := f.Reader(bytes.NewBuffer(out))
+		buf := make([]byte, size)
+		result := []byte{}
+		for n, err := 0, error(nil); err == nil; {
+			n, err = r.Read(buf)
+			result = append(result, buf[:n]...)
+		}
+		return result
+	}
+}
+
+func TestReader(t *testing.T) {
+	for _, s := range bufSizes {
+		name := fmt.Sprintf("TestReader%d", s)
+		runNormTests(t, name, readFunc(s))
+	}
+}
+
+func writeFunc(size int) appendFunc {
+	return func(f Form, out []byte, s string) []byte {
+		in := append(out, s...)
+		result := new(bytes.Buffer)
+		w := f.Writer(result)
+		buf := make([]byte, size)
+		for n := 0; len(in) > 0; in = in[n:] {
+			n = copy(buf, in)
+			_, _ = w.Write(buf[:n])
+		}
+		w.Close()
+		return result.Bytes()
+	}
+}
+
+func TestWriter(t *testing.T) {
+	for _, s := range bufSizes {
+		name := fmt.Sprintf("TestWriter%d", s)
+		runNormTests(t, name, writeFunc(s))
+	}
+}

vendor/golang.org/x/text/unicode/norm/transform_test.go

@@ -0,0 +1,101 @@
+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package norm
+
+import (
+	"fmt"
+	"testing"
+
+	"golang.org/x/text/transform"
+)
+
+func TestTransform(t *testing.T) {
+	tests := []struct {
+		f       Form
+		in, out string
+		eof     bool
+		dstSize int
+		err     error
+	}{
+		{NFC, "ab", "ab", true, 2, nil},
+		{NFC, "qx", "qx", true, 2, nil},
+		{NFD, "qx", "qx", true, 2, nil},
+		{NFC, "", "", true, 1, nil},
+		{NFD, "", "", true, 1, nil},
+		{NFC, "", "", false, 1, nil},
+		{NFD, "", "", false, 1, nil},
+
+		// Normalized segment does not fit in destination.
+		{NFD, "ö", "", true, 1, transform.ErrShortDst},
+		{NFD, "ö", "", true, 2, transform.ErrShortDst},
+
+		// As an artifact of the algorithm, only full segments are written.
+		// This is not strictly required, and some bytes could be written.
+		// In practice, for Transform to not block, the destination buffer
+		// should be at least MaxSegmentSize to work anyway and these edge
+		// conditions will be relatively rare.
+		{NFC, "ab", "", true, 1, transform.ErrShortDst},
+		// This is even true for inert runes.
+		{NFC, "qx", "", true, 1, transform.ErrShortDst},
+		{NFC, "a\u0300abc", "\u00e0a", true, 4, transform.ErrShortDst},
+
+		// We cannot write a segment if succesive runes could still change the result.
+		{NFD, "ö", "", false, 3, transform.ErrShortSrc},
+		{NFC, "a\u0300", "", false, 4, transform.ErrShortSrc},
+		{NFD, "a\u0300", "", false, 4, transform.ErrShortSrc},
+		{NFC, "ö", "", false, 3, transform.ErrShortSrc},
+
+		{NFC, "a\u0300", "", true, 1, transform.ErrShortDst},
+		// Theoretically could fit, but won't due to simplified checks.
+		{NFC, "a\u0300", "", true, 2, transform.ErrShortDst},
+		{NFC, "a\u0300", "", true, 3, transform.ErrShortDst},
+		{NFC, "a\u0300", "\u00e0", true, 4, nil},
+
+		{NFD, "öa\u0300", "o\u0308", false, 8, transform.ErrShortSrc},
+		{NFD, "öa\u0300ö", "o\u0308a\u0300", true, 8, transform.ErrShortDst},
+		{NFD, "öa\u0300ö", "o\u0308a\u0300", false, 12, transform.ErrShortSrc},
+
+		// Illegal input is copied verbatim.
+		{NFD, "\xbd\xb2=\xbc ", "\xbd\xb2=\xbc ", true, 8, nil},
+	}
+	b := make([]byte, 100)
+	for i, tt := range tests {
+		nDst, _, err := tt.f.Transform(b[:tt.dstSize], []byte(tt.in), tt.eof)
+		out := string(b[:nDst])
+		if out != tt.out || err != tt.err {
+			t.Errorf("%d: was %+q (%v); want %+q (%v)", i, out, err, tt.out, tt.err)
+		}
+		if want := tt.f.String(tt.in)[:nDst]; want != out {
+			t.Errorf("%d: incorect normalization: was %+q; want %+q", i, out, want)
+		}
+	}
+}
+
+var transBufSizes = []int{
+	MaxTransformChunkSize,
+	3 * MaxTransformChunkSize / 2,
+	2 * MaxTransformChunkSize,
+	3 * MaxTransformChunkSize,
+	100 * MaxTransformChunkSize,
+}
+
+func doTransNorm(f Form, buf []byte, b []byte) []byte {
+	acc := []byte{}
+	for p := 0; p < len(b); {
+		nd, ns, _ := f.Transform(buf[:], b[p:], true)
+		p += ns
+		acc = append(acc, buf[:nd]...)
+	}
+	return acc
+}
+
+func TestTransformNorm(t *testing.T) {
+	for _, sz := range transBufSizes {
+		buf := make([]byte, sz)
+		runNormTests(t, fmt.Sprintf("Transform:%d", sz), func(f Form, out []byte, s string) []byte {
+			return doTransNorm(f, buf, append(out, s...))
+		})
+	}
+}

vendor/golang.org/x/text/unicode/norm/ucd_test.go

@@ -0,0 +1,275 @@
+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package norm
+
+import (
+	"bufio"
+	"bytes"
+	"fmt"
+	"regexp"
+	"runtime"
+	"strconv"
+	"strings"
+	"sync"
+	"testing"
+	"time"
+	"unicode/utf8"
+
+	"golang.org/x/text/internal/gen"
+	"golang.org/x/text/internal/testtext"
+)
+
+var once sync.Once
+
+func skipShort(t *testing.T) {
+	testtext.SkipIfNotLong(t)
+
+	once.Do(func() { loadTestData(t) })
+}
+
+// This regression test runs the test set in NormalizationTest.txt
+// (taken from http://www.unicode.org/Public/<unicode.Version>/ucd/).
+//
+// NormalizationTest.txt has form:
+// @Part0 # Specific cases
+// #
+// 1E0A;1E0A;0044 0307;1E0A;0044 0307; # (Ḋ; Ḋ; D◌̇; Ḋ; D◌̇; ) LATIN CAPITAL LETTER D WITH DOT ABOVE
+// 1E0C;1E0C;0044 0323;1E0C;0044 0323; # (Ḍ; Ḍ; D◌̣; Ḍ; D◌̣; ) LATIN CAPITAL LETTER D WITH DOT BELOW
+//
+// Each test has 5 columns (c1, c2, c3, c4, c5), where
+// (c1, c2, c3, c4, c5) == (c1, NFC(c1), NFD(c1), NFKC(c1), NFKD(c1))
+//
+// CONFORMANCE:
+// 1. The following invariants must be true for all conformant implementations
+//
+//    NFC
+//      c2 ==  NFC(c1) ==  NFC(c2) ==  NFC(c3)
+//      c4 ==  NFC(c4) ==  NFC(c5)
+//
+//    NFD
+//      c3 ==  NFD(c1) ==  NFD(c2) ==  NFD(c3)
+//      c5 ==  NFD(c4) ==  NFD(c5)
+//
+//    NFKC
+//      c4 == NFKC(c1) == NFKC(c2) == NFKC(c3) == NFKC(c4) == NFKC(c5)
+//
+//    NFKD
+//      c5 == NFKD(c1) == NFKD(c2) == NFKD(c3) == NFKD(c4) == NFKD(c5)
+//
+// 2. For every code point X assigned in this version of Unicode that is not
+//    specifically listed in Part 1, the following invariants must be true
+//    for all conformant implementations:
+//
+//      X == NFC(X) == NFD(X) == NFKC(X) == NFKD(X)
+//
+
+// Column types.
+const (
+	cRaw = iota
+	cNFC
+	cNFD
+	cNFKC
+	cNFKD
+	cMaxColumns
+)
+
+// Holds data from NormalizationTest.txt
+var part []Part
+
+type Part struct {
+	name   string
+	number int
+	tests  []Test
+}
+
+type Test struct {
+	name   string
+	partnr int
+	number int
+	r      rune                // used for character by character test
+	cols   [cMaxColumns]string // Each has 5 entries, see below.
+}
+
+func (t Test) Name() string {
+	if t.number < 0 {
+		return part[t.partnr].name
+	}
+	return fmt.Sprintf("%s:%d", part[t.partnr].name, t.number)
+}
+
+var partRe = regexp.MustCompile(`@Part(\d) # (.*)$`)
+var testRe = regexp.MustCompile(`^` + strings.Repeat(`([\dA-F ]+);`, 5) + ` # (.*)$`)
+
+var counter int
+
+// Load the data form NormalizationTest.txt
+func loadTestData(t *testing.T) {
+	f := gen.OpenUCDFile("NormalizationTest.txt")
+	defer f.Close()
+	scanner := bufio.NewScanner(f)
+	for scanner.Scan() {
+		line := scanner.Text()
+		if len(line) == 0 || line[0] == '#' {
+			continue
+		}
+		m := partRe.FindStringSubmatch(line)
+		if m != nil {
+			if len(m) < 3 {
+				t.Fatal("Failed to parse Part: ", line)
+			}
+			i, err := strconv.Atoi(m[1])
+			if err != nil {
+				t.Fatal(err)
+			}
+			name := m[2]
+			part = append(part, Part{name: name[:len(name)-1], number: i})
+			continue
+		}
+		m = testRe.FindStringSubmatch(line)
+		if m == nil || len(m) < 7 {
+			t.Fatalf(`Failed to parse: "%s" result: %#v`, line, m)
+		}
+		test := Test{name: m[6], partnr: len(part) - 1, number: counter}
+		counter++
+		for j := 1; j < len(m)-1; j++ {
+			for _, split := range strings.Split(m[j], " ") {
+				r, err := strconv.ParseUint(split, 16, 64)
+				if err != nil {
+					t.Fatal(err)
+				}
+				if test.r == 0 {
+					// save for CharacterByCharacterTests
+					test.r = rune(r)
+				}
+				var buf [utf8.UTFMax]byte
+				sz := utf8.EncodeRune(buf[:], rune(r))
+				test.cols[j-1] += string(buf[:sz])
+			}
+		}
+		part := &part[len(part)-1]
+		part.tests = append(part.tests, test)
+	}
+	if scanner.Err() != nil {
+		t.Fatal(scanner.Err())
+	}
+}
+
+func cmpResult(t *testing.T, tc *Test, name string, f Form, gold, test, result string) {
+	if gold != result {
+		t.Errorf("%s:%s: %s(%+q)=%+q; want %+q: %s",
+			tc.Name(), name, fstr[f], test, result, gold, tc.name)
+	}
+}
+
+func cmpIsNormal(t *testing.T, tc *Test, name string, f Form, test string, result, want bool) {
+	if result != want {
+		t.Errorf("%s:%s: %s(%+q)=%v; want %v", tc.Name(), name, fstr[f], test, result, want)
+	}
+}
+
+func doTest(t *testing.T, tc *Test, f Form, gold, test string) {
+	testb := []byte(test)
+	result := f.Bytes(testb)
+	cmpResult(t, tc, "Bytes", f, gold, test, string(result))
+
+	sresult := f.String(test)
+	cmpResult(t, tc, "String", f, gold, test, sresult)
+
+	acc := []byte{}
+	i := Iter{}
+	i.InitString(f, test)
+	for !i.Done() {
+		acc = append(acc, i.Next()...)
+	}
+	cmpResult(t, tc, "Iter.Next", f, gold, test, string(acc))
+
+	buf := make([]byte, 128)
+	acc = nil
+	for p := 0; p < len(testb); {
+		nDst, nSrc, _ := f.Transform(buf, testb[p:], true)
+		acc = append(acc, buf[:nDst]...)
+		p += nSrc
+	}
+	cmpResult(t, tc, "Transform", f, gold, test, string(acc))
+
+	for i := range test {
+		out := f.Append(f.Bytes([]byte(test[:i])), []byte(test[i:])...)
+		cmpResult(t, tc, fmt.Sprintf(":Append:%d", i), f, gold, test, string(out))
+	}
+	cmpIsNormal(t, tc, "IsNormal", f, test, f.IsNormal([]byte(test)), test == gold)
+	cmpIsNormal(t, tc, "IsNormalString", f, test, f.IsNormalString(test), test == gold)
+}
+
+func doConformanceTests(t *testing.T, tc *Test, partn int) {
+	for i := 0; i <= 2; i++ {
+		doTest(t, tc, NFC, tc.cols[1], tc.cols[i])
+		doTest(t, tc, NFD, tc.cols[2], tc.cols[i])
+		doTest(t, tc, NFKC, tc.cols[3], tc.cols[i])
+		doTest(t, tc, NFKD, tc.cols[4], tc.cols[i])
+	}
+	for i := 3; i <= 4; i++ {
+		doTest(t, tc, NFC, tc.cols[3], tc.cols[i])
+		doTest(t, tc, NFD, tc.cols[4], tc.cols[i])
+		doTest(t, tc, NFKC, tc.cols[3], tc.cols[i])
+		doTest(t, tc, NFKD, tc.cols[4], tc.cols[i])
+	}
+}
+
+func TestCharacterByCharacter(t *testing.T) {
+	skipShort(t)
+	tests := part[1].tests
+	var last rune = 0
+	for i := 0; i <= len(tests); i++ { // last one is special case
+		var r rune
+		if i == len(tests) {
+			r = 0x2FA1E // Don't have to go to 0x10FFFF
+		} else {
+			r = tests[i].r
+		}
+		for last++; last < r; last++ {
+			// Check all characters that were not explicitly listed in the test.
+			tc := &Test{partnr: 1, number: -1}
+			char := string(last)
+			doTest(t, tc, NFC, char, char)
+			doTest(t, tc, NFD, char, char)
+			doTest(t, tc, NFKC, char, char)
+			doTest(t, tc, NFKD, char, char)
+		}
+		if i < len(tests) {
+			doConformanceTests(t, &tests[i], 1)
+		}
+	}
+}
+
+func TestStandardTests(t *testing.T) {
+	skipShort(t)
+	for _, j := range []int{0, 2, 3} {
+		for _, test := range part[j].tests {
+			doConformanceTests(t, &test, j)
+		}
+	}
+}
+
+// TestPerformance verifies that normalization is O(n). If any of the
+// code does not properly check for maxCombiningChars, normalization
+// may exhibit O(n**2) behavior.
+func TestPerformance(t *testing.T) {
+	skipShort(t)
+	runtime.GOMAXPROCS(2)
+	success := make(chan bool, 1)
+	go func() {
+		buf := bytes.Repeat([]byte("\u035D"), 1024*1024)
+		buf = append(buf, "\u035B"...)
+		NFC.Append(nil, buf...)
+		success <- true
+	}()
+	timeout := time.After(1 * time.Second)
+	select {
+	case <-success:
+		// test completed before the timeout
+	case <-timeout:
+		t.Errorf(`unexpectedly long time to complete PerformanceTest`)
+	}
+}