diff --git a/.travis.yml b/.travis.yml
index c68d2b0fd..5272b167e 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -7,7 +7,7 @@ services:
 language: go
 
 go:
-  -  1.8.x
+  -  1.9.x
 
 branches:
   only:
diff --git a/README.md b/README.md
index 9f80e8078..8b67e2d74 100644
--- a/README.md
+++ b/README.md
@@ -58,7 +58,7 @@ Developing Nomad
 
 If you wish to work on Nomad itself or any of its built-in systems,
 you will first need [Go](https://www.golang.org) installed on your
-machine (version 1.8+ is *required*).
+machine (version 1.9+ is *required*).
 
 **Developing with Vagrant**
 There is an included Vagrantfile that can help bootstrap the process. The
diff --git a/Vagrantfile b/Vagrantfile
index 029b10b6e..5928f978c 100644
--- a/Vagrantfile
+++ b/Vagrantfile
@@ -6,7 +6,7 @@ VAGRANTFILE_API_VERSION = "2"
 
 DEFAULT_CPU_COUNT = 2
 $script = <<SCRIPT
-GO_VERSION="1.8.3"
+GO_VERSION="1.9.0"
 
 export DEBIAN_FRONTEND=noninteractive
 
diff --git a/nomad/fsm_test.go b/nomad/fsm_test.go
index fa86004d7..b5e60a5e5 100644
--- a/nomad/fsm_test.go
+++ b/nomad/fsm_test.go
@@ -8,6 +8,7 @@ import (
 	"testing"
 	"time"
 
+	"github.com/google/go-cmp/cmp"
 	memdb "github.com/hashicorp/go-memdb"
 	"github.com/hashicorp/nomad/helper"
 	"github.com/hashicorp/nomad/nomad/mock"
@@ -1745,11 +1746,12 @@ func TestFSM_SnapshotRestore_PeriodicLaunches(t *testing.T) {
 	ws := memdb.NewWatchSet()
 	out1, _ := state2.PeriodicLaunchByID(ws, launch1.ID)
 	out2, _ := state2.PeriodicLaunchByID(ws, launch2.ID)
-	if !reflect.DeepEqual(launch1, out1) {
-		t.Fatalf("bad: \n%#v\n%#v", out1, job1)
+
+	if !cmp.Equal(launch1, out1) {
+		t.Fatalf("bad: %v", cmp.Diff(launch1, out1))
 	}
-	if !reflect.DeepEqual(launch2, out2) {
-		t.Fatalf("bad: \n%#v\n%#v", out2, job2)
+	if !cmp.Equal(launch2, out2) {
+		t.Fatalf("bad: %v", cmp.Diff(launch2, out2))
 	}
 }
 
diff --git a/nomad/timetable_test.go b/nomad/timetable_test.go
index 6f51c3b96..00d9a2639 100644
--- a/nomad/timetable_test.go
+++ b/nomad/timetable_test.go
@@ -2,10 +2,12 @@ package nomad
 
 import (
 	"bytes"
-	"reflect"
+	"sync"
 	"testing"
 	"time"
 
+	"github.com/google/go-cmp/cmp"
+	"github.com/google/go-cmp/cmp/cmpopts"
 	"github.com/hashicorp/nomad/nomad/structs"
 	"github.com/ugorji/go/codec"
 )
@@ -122,8 +124,10 @@ func TestTimeTable_SerializeDeserialize(t *testing.T) {
 		t.Fatalf("err: %v", err)
 	}
 
-	if !reflect.DeepEqual(tt.table, tt2.table) {
-		t.Fatalf("bad: %#v %#v", tt, tt2)
+	o := cmp.AllowUnexported(TimeTable{})
+	o2 := cmpopts.IgnoreTypes(sync.RWMutex{})
+	if !cmp.Equal(tt.table, tt2.table, o, o2) {
+		t.Fatalf("bad: %s", cmp.Diff(tt, tt2, o, o2))
 	}
 }
 
diff --git a/vendor/github.com/google/go-cmp/LICENSE b/vendor/github.com/google/go-cmp/LICENSE
new file mode 100644
index 000000000..32017f8fa
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/LICENSE
@@ -0,0 +1,27 @@
+Copyright (c) 2017 The Go Authors. All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+
+   * Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+   * Redistributions in binary form must reproduce the above
+copyright notice, this list of conditions and the following disclaimer
+in the documentation and/or other materials provided with the
+distribution.
+   * Neither the name of Google Inc. nor the names of its
+contributors may be used to endorse or promote products derived from
+this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
diff --git a/vendor/github.com/google/go-cmp/cmp/cmpopts/equate.go b/vendor/github.com/google/go-cmp/cmp/cmpopts/equate.go
new file mode 100644
index 000000000..cc39492cf
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/cmpopts/equate.go
@@ -0,0 +1,89 @@
+// Copyright 2017, 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.md file.
+
+// Package cmpopts provides common options for the cmp package.
+package cmpopts
+
+import (
+	"math"
+	"reflect"
+
+	"github.com/google/go-cmp/cmp"
+)
+
+func equateAlways(_, _ interface{}) bool { return true }
+
+// EquateEmpty returns a Comparer option that determines all maps and slices
+// with a length of zero to be equal, regardless of whether they are nil.
+//
+// EquateEmpty can be used in conjuction with SortSlices and SortMaps.
+func EquateEmpty() cmp.Option {
+	return cmp.FilterValues(isEmpty, cmp.Comparer(equateAlways))
+}
+
+func isEmpty(x, y interface{}) bool {
+	vx, vy := reflect.ValueOf(x), reflect.ValueOf(y)
+	return (x != nil && y != nil && vx.Type() == vy.Type()) &&
+		(vx.Kind() == reflect.Slice || vx.Kind() == reflect.Map) &&
+		(vx.Len() == 0 && vy.Len() == 0)
+}
+
+// EquateApprox returns a Comparer option that determines float32 or float64
+// values to be equal if they are within a relative fraction or absolute margin.
+// This option is not used when either x or y is NaN or infinite.
+//
+// The fraction determines that the difference of two values must be within the
+// smaller fraction of the two values, while the margin determines that the two
+// values must be within some absolute margin.
+// To express only a fraction or only a margin, use 0 for the other parameter.
+// The fraction and margin must be non-negative.
+//
+// The mathematical expression used is equivalent to:
+//	|x-y| ≤ max(fraction*min(|x|, |y|), margin)
+//
+// EquateApprox can be used in conjuction with EquateNaNs.
+func EquateApprox(fraction, margin float64) cmp.Option {
+	if margin < 0 || fraction < 0 || math.IsNaN(margin) || math.IsNaN(fraction) {
+		panic("margin or fraction must be a non-negative number")
+	}
+	a := approximator{fraction, margin}
+	return cmp.Options{
+		cmp.FilterValues(areRealF64s, cmp.Comparer(a.compareF64)),
+		cmp.FilterValues(areRealF32s, cmp.Comparer(a.compareF32)),
+	}
+}
+
+type approximator struct{ frac, marg float64 }
+
+func areRealF64s(x, y float64) bool {
+	return !math.IsNaN(x) && !math.IsNaN(y) && !math.IsInf(x, 0) && !math.IsInf(y, 0)
+}
+func areRealF32s(x, y float32) bool {
+	return areRealF64s(float64(x), float64(y))
+}
+func (a approximator) compareF64(x, y float64) bool {
+	relMarg := a.frac * math.Min(math.Abs(x), math.Abs(y))
+	return math.Abs(x-y) <= math.Max(a.marg, relMarg)
+}
+func (a approximator) compareF32(x, y float32) bool {
+	return a.compareF64(float64(x), float64(y))
+}
+
+// EquateNaNs returns a Comparer option that determines float32 and float64
+// NaN values to be equal.
+//
+// EquateNaNs can be used in conjuction with EquateApprox.
+func EquateNaNs() cmp.Option {
+	return cmp.Options{
+		cmp.FilterValues(areNaNsF64s, cmp.Comparer(equateAlways)),
+		cmp.FilterValues(areNaNsF32s, cmp.Comparer(equateAlways)),
+	}
+}
+
+func areNaNsF64s(x, y float64) bool {
+	return math.IsNaN(x) && math.IsNaN(y)
+}
+func areNaNsF32s(x, y float32) bool {
+	return areNaNsF64s(float64(x), float64(y))
+}
diff --git a/vendor/github.com/google/go-cmp/cmp/cmpopts/ignore.go b/vendor/github.com/google/go-cmp/cmp/cmpopts/ignore.go
new file mode 100644
index 000000000..016891da3
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/cmpopts/ignore.go
@@ -0,0 +1,148 @@
+// Copyright 2017, 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.md file.
+
+package cmpopts
+
+import (
+	"fmt"
+	"reflect"
+	"unicode"
+	"unicode/utf8"
+
+	"github.com/google/go-cmp/cmp"
+)
+
+// IgnoreFields returns an Option that ignores exported fields of the
+// given names on a single struct type.
+// The struct type is specified by passing in a value of that type.
+//
+// The name may be a dot-delimited string (e.g., "Foo.Bar") to ignore a
+// specific sub-field that is embedded or nested within the parent struct.
+//
+// This does not handle unexported fields; use IgnoreUnexported instead.
+func IgnoreFields(typ interface{}, names ...string) cmp.Option {
+	sf := newStructFilter(typ, names...)
+	return cmp.FilterPath(sf.filter, cmp.Ignore())
+}
+
+// IgnoreTypes returns an Option that ignores all values assignable to
+// certain types, which are specified by passing in a value of each type.
+func IgnoreTypes(typs ...interface{}) cmp.Option {
+	tf := newTypeFilter(typs...)
+	return cmp.FilterPath(tf.filter, cmp.Ignore())
+}
+
+type typeFilter []reflect.Type
+
+func newTypeFilter(typs ...interface{}) (tf typeFilter) {
+	for _, typ := range typs {
+		t := reflect.TypeOf(typ)
+		if t == nil {
+			// This occurs if someone tries to pass in sync.Locker(nil)
+			panic("cannot determine type; consider using IgnoreInterfaces")
+		}
+		tf = append(tf, t)
+	}
+	return tf
+}
+func (tf typeFilter) filter(p cmp.Path) bool {
+	if len(p) < 1 {
+		return false
+	}
+	t := p[len(p)-1].Type()
+	for _, ti := range tf {
+		if t.AssignableTo(ti) {
+			return true
+		}
+	}
+	return false
+}
+
+// IgnoreInterfaces returns an Option that ignores all values or references of
+// values assignable to certain interface types. These interfaces are specified
+// by passing in an anonymous struct with the interface types embedded in it.
+// For example, to ignore sync.Locker, pass in struct{sync.Locker}{}.
+func IgnoreInterfaces(ifaces interface{}) cmp.Option {
+	tf := newIfaceFilter(ifaces)
+	return cmp.FilterPath(tf.filter, cmp.Ignore())
+}
+
+type ifaceFilter []reflect.Type
+
+func newIfaceFilter(ifaces interface{}) (tf ifaceFilter) {
+	t := reflect.TypeOf(ifaces)
+	if ifaces == nil || t.Name() != "" || t.Kind() != reflect.Struct {
+		panic("input must be an anonymous struct")
+	}
+	for i := 0; i < t.NumField(); i++ {
+		fi := t.Field(i)
+		switch {
+		case !fi.Anonymous:
+			panic("struct cannot have named fields")
+		case fi.Type.Kind() != reflect.Interface:
+			panic("embedded field must be an interface type")
+		case fi.Type.NumMethod() == 0:
+			// This matches everything; why would you ever want this?
+			panic("cannot ignore empty interface")
+		default:
+			tf = append(tf, fi.Type)
+		}
+	}
+	return tf
+}
+func (tf ifaceFilter) filter(p cmp.Path) bool {
+	if len(p) < 1 {
+		return false
+	}
+	t := p[len(p)-1].Type()
+	for _, ti := range tf {
+		if t.AssignableTo(ti) {
+			return true
+		}
+		if t.Kind() != reflect.Ptr && reflect.PtrTo(t).AssignableTo(ti) {
+			return true
+		}
+	}
+	return false
+}
+
+// IgnoreUnexported returns an Option that only ignores the immediate unexported
+// fields of a struct, including anonymous fields of unexported types.
+// In particular, unexported fields within the struct's exported fields
+// of struct types, including anonymous fields, will not be ignored unless the
+// type of the field itself is also passed to IgnoreUnexported.
+func IgnoreUnexported(typs ...interface{}) cmp.Option {
+	ux := newUnexportedFilter(typs...)
+	return cmp.FilterPath(ux.filter, cmp.Ignore())
+}
+
+type unexportedFilter struct{ m map[reflect.Type]bool }
+
+func newUnexportedFilter(typs ...interface{}) unexportedFilter {
+	ux := unexportedFilter{m: make(map[reflect.Type]bool)}
+	for _, typ := range typs {
+		t := reflect.TypeOf(typ)
+		if t == nil || t.Kind() != reflect.Struct {
+			panic(fmt.Sprintf("invalid struct type: %T", typ))
+		}
+		ux.m[t] = true
+	}
+	return ux
+}
+func (xf unexportedFilter) filter(p cmp.Path) bool {
+	if len(p) < 2 {
+		return false
+	}
+	sf, ok := p[len(p)-1].(cmp.StructField)
+	if !ok {
+		return false
+	}
+	return xf.m[p[len(p)-2].Type()] && !isExported(sf.Name())
+}
+
+// isExported reports whether the identifier is exported.
+func isExported(id string) bool {
+	r, _ := utf8.DecodeRuneInString(id)
+	return unicode.IsUpper(r)
+}
diff --git a/vendor/github.com/google/go-cmp/cmp/cmpopts/sort.go b/vendor/github.com/google/go-cmp/cmp/cmpopts/sort.go
new file mode 100644
index 000000000..a566d240b
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/cmpopts/sort.go
@@ -0,0 +1,146 @@
+// Copyright 2017, 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.md file.
+
+package cmpopts
+
+import (
+	"fmt"
+	"reflect"
+
+	"github.com/google/go-cmp/cmp"
+	"github.com/google/go-cmp/cmp/internal/function"
+)
+
+// SortSlices returns a Transformer option that sorts all []V.
+// The less function must be of the form "func(T, T) bool" which is used to
+// sort any slice with element type V that is assignable to T.
+//
+// The less function must be:
+//	• Deterministic: less(x, y) == less(x, y)
+//	• Irreflexive: !less(x, x)
+//	• Transitive: if !less(x, y) and !less(y, z), then !less(x, z)
+//
+// The less function does not have to be "total". That is, if !less(x, y) and
+// !less(y, x) for two elements x and y, their relative order is maintained.
+//
+// SortSlices can be used in conjuction with EquateEmpty.
+func SortSlices(less interface{}) cmp.Option {
+	vf := reflect.ValueOf(less)
+	if !function.IsType(vf.Type(), function.Less) || vf.IsNil() {
+		panic(fmt.Sprintf("invalid less function: %T", less))
+	}
+	ss := sliceSorter{vf.Type().In(0), vf}
+	return cmp.FilterValues(ss.filter, cmp.Transformer("Sort", ss.sort))
+}
+
+type sliceSorter struct {
+	in  reflect.Type  // T
+	fnc reflect.Value // func(T, T) bool
+}
+
+func (ss sliceSorter) filter(x, y interface{}) bool {
+	vx, vy := reflect.ValueOf(x), reflect.ValueOf(y)
+	if !(x != nil && y != nil && vx.Type() == vy.Type()) ||
+		!(vx.Kind() == reflect.Slice && vx.Type().Elem().AssignableTo(ss.in)) ||
+		(vx.Len() <= 1 && vy.Len() <= 1) {
+		return false
+	}
+	// Check whether the slices are already sorted to avoid an infinite
+	// recursion cycle applying the same transform to itself.
+	ok1 := sliceIsSorted(x, func(i, j int) bool { return ss.less(vx, i, j) })
+	ok2 := sliceIsSorted(y, func(i, j int) bool { return ss.less(vy, i, j) })
+	return !ok1 || !ok2
+}
+func (ss sliceSorter) sort(x interface{}) interface{} {
+	src := reflect.ValueOf(x)
+	dst := reflect.MakeSlice(src.Type(), src.Len(), src.Len())
+	for i := 0; i < src.Len(); i++ {
+		dst.Index(i).Set(src.Index(i))
+	}
+	sortSliceStable(dst.Interface(), func(i, j int) bool { return ss.less(dst, i, j) })
+	ss.checkSort(dst)
+	return dst.Interface()
+}
+func (ss sliceSorter) checkSort(v reflect.Value) {
+	start := -1 // Start of a sequence of equal elements.
+	for i := 1; i < v.Len(); i++ {
+		if ss.less(v, i-1, i) {
+			// Check that first and last elements in v[start:i] are equal.
+			if start >= 0 && (ss.less(v, start, i-1) || ss.less(v, i-1, start)) {
+				panic(fmt.Sprintf("incomparable values detected: want equal elements: %v", v.Slice(start, i)))
+			}
+			start = -1
+		} else if start == -1 {
+			start = i
+		}
+	}
+}
+func (ss sliceSorter) less(v reflect.Value, i, j int) bool {
+	vx, vy := v.Index(i), v.Index(j)
+	return ss.fnc.Call([]reflect.Value{vx, vy})[0].Bool()
+}
+
+// SortMaps returns a Transformer option that flattens map[K]V types to be a
+// sorted []struct{K, V}. The less function must be of the form
+// "func(T, T) bool" which is used to sort any map with key K that is
+// assignable to T.
+//
+// Flattening the map into a slice has the property that cmp.Equal is able to
+// use Comparers on K or the K.Equal method if it exists.
+//
+// The less function must be:
+//	• Deterministic: less(x, y) == less(x, y)
+//	• Irreflexive: !less(x, x)
+//	• Transitive: if !less(x, y) and !less(y, z), then !less(x, z)
+//	• Total: if x != y, then either less(x, y) or less(y, x)
+//
+// SortMaps can be used in conjuction with EquateEmpty.
+func SortMaps(less interface{}) cmp.Option {
+	vf := reflect.ValueOf(less)
+	if !function.IsType(vf.Type(), function.Less) || vf.IsNil() {
+		panic(fmt.Sprintf("invalid less function: %T", less))
+	}
+	ms := mapSorter{vf.Type().In(0), vf}
+	return cmp.FilterValues(ms.filter, cmp.Transformer("Sort", ms.sort))
+}
+
+type mapSorter struct {
+	in  reflect.Type  // T
+	fnc reflect.Value // func(T, T) bool
+}
+
+func (ms mapSorter) filter(x, y interface{}) bool {
+	vx, vy := reflect.ValueOf(x), reflect.ValueOf(y)
+	return (x != nil && y != nil && vx.Type() == vy.Type()) &&
+		(vx.Kind() == reflect.Map && vx.Type().Key().AssignableTo(ms.in)) &&
+		(vx.Len() != 0 || vy.Len() != 0)
+}
+func (ms mapSorter) sort(x interface{}) interface{} {
+	src := reflect.ValueOf(x)
+	outType := mapEntryType(src.Type())
+	dst := reflect.MakeSlice(reflect.SliceOf(outType), src.Len(), src.Len())
+	for i, k := range src.MapKeys() {
+		v := reflect.New(outType).Elem()
+		v.Field(0).Set(k)
+		v.Field(1).Set(src.MapIndex(k))
+		dst.Index(i).Set(v)
+	}
+	sortSlice(dst.Interface(), func(i, j int) bool { return ms.less(dst, i, j) })
+	ms.checkSort(dst)
+	return dst.Interface()
+}
+func (ms mapSorter) checkSort(v reflect.Value) {
+	for i := 1; i < v.Len(); i++ {
+		if !ms.less(v, i-1, i) {
+			panic(fmt.Sprintf("partial order detected: want %v < %v", v.Index(i-1), v.Index(i)))
+		}
+	}
+}
+func (ms mapSorter) less(v reflect.Value, i, j int) bool {
+	vx, vy := v.Index(i).Field(0), v.Index(j).Field(0)
+	if !hasReflectStructOf {
+		vx, vy = vx.Elem(), vy.Elem()
+	}
+	return ms.fnc.Call([]reflect.Value{vx, vy})[0].Bool()
+}
diff --git a/vendor/github.com/google/go-cmp/cmp/cmpopts/sort_go17.go b/vendor/github.com/google/go-cmp/cmp/cmpopts/sort_go17.go
new file mode 100644
index 000000000..839b88ca4
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/cmpopts/sort_go17.go
@@ -0,0 +1,46 @@
+// Copyright 2017, 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.md file.
+
+// +build !go1.8
+
+package cmpopts
+
+import (
+	"reflect"
+	"sort"
+)
+
+const hasReflectStructOf = false
+
+func mapEntryType(reflect.Type) reflect.Type {
+	return reflect.TypeOf(struct{ K, V interface{} }{})
+}
+
+func sliceIsSorted(slice interface{}, less func(i, j int) bool) bool {
+	return sort.IsSorted(reflectSliceSorter{reflect.ValueOf(slice), less})
+}
+func sortSlice(slice interface{}, less func(i, j int) bool) {
+	sort.Sort(reflectSliceSorter{reflect.ValueOf(slice), less})
+}
+func sortSliceStable(slice interface{}, less func(i, j int) bool) {
+	sort.Stable(reflectSliceSorter{reflect.ValueOf(slice), less})
+}
+
+type reflectSliceSorter struct {
+	slice reflect.Value
+	less  func(i, j int) bool
+}
+
+func (ss reflectSliceSorter) Len() int {
+	return ss.slice.Len()
+}
+func (ss reflectSliceSorter) Less(i, j int) bool {
+	return ss.less(i, j)
+}
+func (ss reflectSliceSorter) Swap(i, j int) {
+	vi := ss.slice.Index(i).Interface()
+	vj := ss.slice.Index(j).Interface()
+	ss.slice.Index(i).Set(reflect.ValueOf(vj))
+	ss.slice.Index(j).Set(reflect.ValueOf(vi))
+}
diff --git a/vendor/github.com/google/go-cmp/cmp/cmpopts/sort_go18.go b/vendor/github.com/google/go-cmp/cmp/cmpopts/sort_go18.go
new file mode 100644
index 000000000..8a59c0d38
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/cmpopts/sort_go18.go
@@ -0,0 +1,31 @@
+// Copyright 2017, 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.md file.
+
+// +build go1.8
+
+package cmpopts
+
+import (
+	"reflect"
+	"sort"
+)
+
+const hasReflectStructOf = true
+
+func mapEntryType(t reflect.Type) reflect.Type {
+	return reflect.StructOf([]reflect.StructField{
+		{Name: "K", Type: t.Key()},
+		{Name: "V", Type: t.Elem()},
+	})
+}
+
+func sliceIsSorted(slice interface{}, less func(i, j int) bool) bool {
+	return sort.SliceIsSorted(slice, less)
+}
+func sortSlice(slice interface{}, less func(i, j int) bool) {
+	sort.Slice(slice, less)
+}
+func sortSliceStable(slice interface{}, less func(i, j int) bool) {
+	sort.SliceStable(slice, less)
+}
diff --git a/vendor/github.com/google/go-cmp/cmp/cmpopts/struct_filter.go b/vendor/github.com/google/go-cmp/cmp/cmpopts/struct_filter.go
new file mode 100644
index 000000000..97f707983
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/cmpopts/struct_filter.go
@@ -0,0 +1,182 @@
+// Copyright 2017, 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.md file.
+
+package cmpopts
+
+import (
+	"fmt"
+	"reflect"
+	"strings"
+
+	"github.com/google/go-cmp/cmp"
+)
+
+// filterField returns a new Option where opt is only evaluated on paths that
+// include a specific exported field on a single struct type.
+// The struct type is specified by passing in a value of that type.
+//
+// The name may be a dot-delimited string (e.g., "Foo.Bar") to select a
+// specific sub-field that is embedded or nested within the parent struct.
+func filterField(typ interface{}, name string, opt cmp.Option) cmp.Option {
+	// TODO: This is currently unexported over concerns of how helper filters
+	// can be composed together easily.
+	// TODO: Add tests for FilterField.
+
+	sf := newStructFilter(typ, name)
+	return cmp.FilterPath(sf.filter, opt)
+}
+
+type structFilter struct {
+	t  reflect.Type // The root struct type to match on
+	ft fieldTree    // Tree of fields to match on
+}
+
+func newStructFilter(typ interface{}, names ...string) structFilter {
+	// TODO: Perhaps allow * as a special identifier to allow ignoring any
+	// number of path steps until the next field match?
+	// This could be useful when a concrete struct gets transformed into
+	// an anonymous struct where it is not possible to specify that by type,
+	// but the transformer happens to provide guarantees about the names of
+	// the transformed fields.
+
+	t := reflect.TypeOf(typ)
+	if t == nil || t.Kind() != reflect.Struct {
+		panic(fmt.Sprintf("%T must be a struct", typ))
+	}
+	var ft fieldTree
+	for _, name := range names {
+		cname, err := canonicalName(t, name)
+		if err != nil {
+			panic(fmt.Sprintf("%s: %v", strings.Join(cname, "."), err))
+		}
+		ft.insert(cname)
+	}
+	return structFilter{t, ft}
+}
+
+func (sf structFilter) filter(p cmp.Path) bool {
+	for i, ps := range p {
+		if ps.Type().AssignableTo(sf.t) && sf.ft.matchPrefix(p[i+1:]) {
+			return true
+		}
+	}
+	return false
+}
+
+// fieldTree represents a set of dot-separated identifiers.
+//
+// For example, inserting the following selectors:
+//	Foo
+//	Foo.Bar.Baz
+//	Foo.Buzz
+//	Nuka.Cola.Quantum
+//
+// Results in a tree of the form:
+//	{sub: {
+//		"Foo": {ok: true, sub: {
+//			"Bar": {sub: {
+//				"Baz": {ok: true},
+//			}},
+//			"Buzz": {ok: true},
+//		}},
+//		"Nuka": {sub: {
+//			"Cola": {sub: {
+//				"Quantum": {ok: true},
+//			}},
+//		}},
+//	}}
+type fieldTree struct {
+	ok  bool                 // Whether this is a specified node
+	sub map[string]fieldTree // The sub-tree of fields under this node
+}
+
+// insert inserts a sequence of field accesses into the tree.
+func (ft *fieldTree) insert(cname []string) {
+	if ft.sub == nil {
+		ft.sub = make(map[string]fieldTree)
+	}
+	if len(cname) == 0 {
+		ft.ok = true
+		return
+	}
+	sub := ft.sub[cname[0]]
+	sub.insert(cname[1:])
+	ft.sub[cname[0]] = sub
+}
+
+// matchPrefix reports whether any selector in the fieldTree matches
+// the start of path p.
+func (ft fieldTree) matchPrefix(p cmp.Path) bool {
+	for _, ps := range p {
+		switch ps := ps.(type) {
+		case cmp.StructField:
+			ft = ft.sub[ps.Name()]
+			if ft.ok {
+				return true
+			}
+			if len(ft.sub) == 0 {
+				return false
+			}
+		case cmp.Indirect:
+		default:
+			return false
+		}
+	}
+	return false
+}
+
+// canonicalName returns a list of identifiers where any struct field access
+// through an embedded field is expanded to include the names of the embedded
+// types themselves.
+//
+// For example, suppose field "Foo" is not directly in the parent struct,
+// but actually from an embedded struct of type "Bar". Then, the canonical name
+// of "Foo" is actually "Bar.Foo".
+//
+// Suppose field "Foo" is not directly in the parent struct, but actually
+// a field in two different embedded structs of types "Bar" and "Baz".
+// Then the selector "Foo" causes a panic since it is ambiguous which one it
+// refers to. The user must specify either "Bar.Foo" or "Baz.Foo".
+func canonicalName(t reflect.Type, sel string) ([]string, error) {
+	var name string
+	sel = strings.TrimPrefix(sel, ".")
+	if sel == "" {
+		return nil, fmt.Errorf("name must not be empty")
+	}
+	if i := strings.IndexByte(sel, '.'); i < 0 {
+		name, sel = sel, ""
+	} else {
+		name, sel = sel[:i], sel[i:]
+	}
+
+	// Type must be a struct or pointer to struct.
+	if t.Kind() == reflect.Ptr {
+		t = t.Elem()
+	}
+	if t.Kind() != reflect.Struct {
+		return nil, fmt.Errorf("%v must be a struct", t)
+	}
+
+	// Find the canonical name for this current field name.
+	// If the field exists in an embedded struct, then it will be expanded.
+	if !isExported(name) {
+		// Disallow unexported fields:
+		//	* To discourage people from actually touching unexported fields
+		//	* FieldByName is buggy (https://golang.org/issue/4876)
+		return []string{name}, fmt.Errorf("name must be exported")
+	}
+	sf, ok := t.FieldByName(name)
+	if !ok {
+		return []string{name}, fmt.Errorf("does not exist")
+	}
+	var ss []string
+	for i := range sf.Index {
+		ss = append(ss, t.FieldByIndex(sf.Index[:i+1]).Name)
+	}
+	if sel == "" {
+		return ss, nil
+	}
+	ssPost, err := canonicalName(sf.Type, sel)
+	return append(ss, ssPost...), err
+}
diff --git a/vendor/github.com/google/go-cmp/cmp/compare.go b/vendor/github.com/google/go-cmp/cmp/compare.go
new file mode 100644
index 000000000..5527f0149
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/compare.go
@@ -0,0 +1,529 @@
+// Copyright 2017, 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.md file.
+
+// Package cmp determines equality of values.
+//
+// This package is intended to be a more powerful and safer alternative to
+// reflect.DeepEqual for comparing whether two values are semantically equal.
+//
+// The primary features of cmp are:
+//
+// • When the default behavior of equality does not suit the needs of the test,
+// custom equality functions can override the equality operation.
+// For example, an equality function may report floats as equal so long as they
+// are within some tolerance of each other.
+//
+// • Types that have an Equal method may use that method to determine equality.
+// This allows package authors to determine the equality operation for the types
+// that they define.
+//
+// • If no custom equality functions are used and no Equal method is defined,
+// equality is determined by recursively comparing the primitive kinds on both
+// values, much like reflect.DeepEqual. Unlike reflect.DeepEqual, unexported
+// fields are not compared by default; they result in panics unless suppressed
+// by using an Ignore option (see cmpopts.IgnoreUnexported) or explictly compared
+// using the AllowUnexported option.
+package cmp
+
+import (
+	"fmt"
+	"reflect"
+
+	"github.com/google/go-cmp/cmp/internal/diff"
+	"github.com/google/go-cmp/cmp/internal/function"
+	"github.com/google/go-cmp/cmp/internal/value"
+)
+
+// BUG: Maps with keys containing NaN values cannot be properly compared due to
+// the reflection package's inability to retrieve such entries. Equal will panic
+// anytime it comes across a NaN key, but this behavior may change.
+//
+// See https://golang.org/issue/11104 for more details.
+
+var nothing = reflect.Value{}
+
+// Equal reports whether x and y are equal by recursively applying the
+// following rules in the given order to x and y and all of their sub-values:
+//
+// • If two values are not of the same type, then they are never equal
+// and the overall result is false.
+//
+// • Let S be the set of all Ignore, Transformer, and Comparer options that
+// remain after applying all path filters, value filters, and type filters.
+// If at least one Ignore exists in S, then the comparison is ignored.
+// If the number of Transformer and Comparer options in S is greater than one,
+// then Equal panics because it is ambiguous which option to use.
+// If S contains a single Transformer, then use that to transform the current
+// values and recursively call Equal on the output values.
+// If S contains a single Comparer, then use that to compare the current values.
+// Otherwise, evaluation proceeds to the next rule.
+//
+// • If the values have an Equal method of the form "(T) Equal(T) bool" or
+// "(T) Equal(I) bool" where T is assignable to I, then use the result of
+// x.Equal(y). Otherwise, no such method exists and evaluation proceeds to
+// the next rule.
+//
+// • Lastly, try to compare x and y based on their basic kinds.
+// Simple kinds like booleans, integers, floats, complex numbers, strings, and
+// channels are compared using the equivalent of the == operator in Go.
+// Functions are only equal if they are both nil, otherwise they are unequal.
+// Pointers are equal if the underlying values they point to are also equal.
+// Interfaces are equal if their underlying concrete values are also equal.
+//
+// Structs are equal if all of their fields are equal. If a struct contains
+// unexported fields, Equal panics unless the AllowUnexported option is used or
+// an Ignore option (e.g., cmpopts.IgnoreUnexported) ignores that field.
+//
+// Arrays, slices, and maps are equal if they are both nil or both non-nil
+// with the same length and the elements at each index or key are equal.
+// Note that a non-nil empty slice and a nil slice are not equal.
+// To equate empty slices and maps, consider using cmpopts.EquateEmpty.
+// Map keys are equal according to the == operator.
+// To use custom comparisons for map keys, consider using cmpopts.SortMaps.
+func Equal(x, y interface{}, opts ...Option) bool {
+	s := newState(opts)
+	s.compareAny(reflect.ValueOf(x), reflect.ValueOf(y))
+	return s.result.Equal()
+}
+
+// Diff returns a human-readable report of the differences between two values.
+// It returns an empty string if and only if Equal returns true for the same
+// input values and options. The output string will use the "-" symbol to
+// indicate elements removed from x, and the "+" symbol to indicate elements
+// added to y.
+//
+// Do not depend on this output being stable.
+func Diff(x, y interface{}, opts ...Option) string {
+	r := new(defaultReporter)
+	opts = Options{Options(opts), r}
+	eq := Equal(x, y, opts...)
+	d := r.String()
+	if (d == "") != eq {
+		panic("inconsistent difference and equality results")
+	}
+	return d
+}
+
+type state struct {
+	// These fields represent the "comparison state".
+	// Calling statelessCompare must not result in observable changes to these.
+	result   diff.Result // The current result of comparison
+	curPath  Path        // The current path in the value tree
+	reporter reporter    // Optional reporter used for difference formatting
+
+	// dynChecker triggers pseudo-random checks for option correctness.
+	// It is safe for statelessCompare to mutate this value.
+	dynChecker dynChecker
+
+	// These fields, once set by processOption, will not change.
+	exporters map[reflect.Type]bool // Set of structs with unexported field visibility
+	opts      Options               // List of all fundamental and filter options
+}
+
+func newState(opts []Option) *state {
+	s := new(state)
+	for _, opt := range opts {
+		s.processOption(opt)
+	}
+	return s
+}
+
+func (s *state) processOption(opt Option) {
+	switch opt := opt.(type) {
+	case nil:
+	case Options:
+		for _, o := range opt {
+			s.processOption(o)
+		}
+	case coreOption:
+		type filtered interface {
+			isFiltered() bool
+		}
+		if fopt, ok := opt.(filtered); ok && !fopt.isFiltered() {
+			panic(fmt.Sprintf("cannot use an unfiltered option: %v", opt))
+		}
+		s.opts = append(s.opts, opt)
+	case visibleStructs:
+		if s.exporters == nil {
+			s.exporters = make(map[reflect.Type]bool)
+		}
+		for t := range opt {
+			s.exporters[t] = true
+		}
+	case reporter:
+		if s.reporter != nil {
+			panic("difference reporter already registered")
+		}
+		s.reporter = opt
+	default:
+		panic(fmt.Sprintf("unknown option %T", opt))
+	}
+}
+
+// statelessCompare compares two values and returns the result.
+// This function is stateless in that it does not alter the current result,
+// or output to any registered reporters.
+func (s *state) statelessCompare(vx, vy reflect.Value) diff.Result {
+	// We do not save and restore the curPath because all of the compareX
+	// methods should properly push and pop from the path.
+	// It is an implementation bug if the contents of curPath differs from
+	// when calling this function to when returning from it.
+
+	oldResult, oldReporter := s.result, s.reporter
+	s.result = diff.Result{} // Reset result
+	s.reporter = nil         // Remove reporter to avoid spurious printouts
+	s.compareAny(vx, vy)
+	res := s.result
+	s.result, s.reporter = oldResult, oldReporter
+	return res
+}
+
+func (s *state) compareAny(vx, vy reflect.Value) {
+	// TODO: Support cyclic data structures.
+
+	// Rule 0: Differing types are never equal.
+	if !vx.IsValid() || !vy.IsValid() {
+		s.report(vx.IsValid() == vy.IsValid(), vx, vy)
+		return
+	}
+	if vx.Type() != vy.Type() {
+		s.report(false, vx, vy) // Possible for path to be empty
+		return
+	}
+	t := vx.Type()
+	if len(s.curPath) == 0 {
+		s.curPath.push(&pathStep{typ: t})
+		defer s.curPath.pop()
+	}
+	vx, vy = s.tryExporting(vx, vy)
+
+	// Rule 1: Check whether an option applies on this node in the value tree.
+	if s.tryOptions(vx, vy, t) {
+		return
+	}
+
+	// Rule 2: Check whether the type has a valid Equal method.
+	if s.tryMethod(vx, vy, t) {
+		return
+	}
+
+	// Rule 3: Recursively descend into each value's underlying kind.
+	switch t.Kind() {
+	case reflect.Bool:
+		s.report(vx.Bool() == vy.Bool(), vx, vy)
+		return
+	case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64:
+		s.report(vx.Int() == vy.Int(), vx, vy)
+		return
+	case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr:
+		s.report(vx.Uint() == vy.Uint(), vx, vy)
+		return
+	case reflect.Float32, reflect.Float64:
+		s.report(vx.Float() == vy.Float(), vx, vy)
+		return
+	case reflect.Complex64, reflect.Complex128:
+		s.report(vx.Complex() == vy.Complex(), vx, vy)
+		return
+	case reflect.String:
+		s.report(vx.String() == vy.String(), vx, vy)
+		return
+	case reflect.Chan, reflect.UnsafePointer:
+		s.report(vx.Pointer() == vy.Pointer(), vx, vy)
+		return
+	case reflect.Func:
+		s.report(vx.IsNil() && vy.IsNil(), vx, vy)
+		return
+	case reflect.Ptr:
+		if vx.IsNil() || vy.IsNil() {
+			s.report(vx.IsNil() && vy.IsNil(), vx, vy)
+			return
+		}
+		s.curPath.push(&indirect{pathStep{t.Elem()}})
+		defer s.curPath.pop()
+		s.compareAny(vx.Elem(), vy.Elem())
+		return
+	case reflect.Interface:
+		if vx.IsNil() || vy.IsNil() {
+			s.report(vx.IsNil() && vy.IsNil(), vx, vy)
+			return
+		}
+		if vx.Elem().Type() != vy.Elem().Type() {
+			s.report(false, vx.Elem(), vy.Elem())
+			return
+		}
+		s.curPath.push(&typeAssertion{pathStep{vx.Elem().Type()}})
+		defer s.curPath.pop()
+		s.compareAny(vx.Elem(), vy.Elem())
+		return
+	case reflect.Slice:
+		if vx.IsNil() || vy.IsNil() {
+			s.report(vx.IsNil() && vy.IsNil(), vx, vy)
+			return
+		}
+		fallthrough
+	case reflect.Array:
+		s.compareArray(vx, vy, t)
+		return
+	case reflect.Map:
+		s.compareMap(vx, vy, t)
+		return
+	case reflect.Struct:
+		s.compareStruct(vx, vy, t)
+		return
+	default:
+		panic(fmt.Sprintf("%v kind not handled", t.Kind()))
+	}
+}
+
+func (s *state) tryExporting(vx, vy reflect.Value) (reflect.Value, reflect.Value) {
+	if sf, ok := s.curPath[len(s.curPath)-1].(*structField); ok && sf.unexported {
+		if sf.force {
+			// Use unsafe pointer arithmetic to get read-write access to an
+			// unexported field in the struct.
+			vx = unsafeRetrieveField(sf.pvx, sf.field)
+			vy = unsafeRetrieveField(sf.pvy, sf.field)
+		} else {
+			// We are not allowed to export the value, so invalidate them
+			// so that tryOptions can panic later if not explicitly ignored.
+			vx = nothing
+			vy = nothing
+		}
+	}
+	return vx, vy
+}
+
+func (s *state) tryOptions(vx, vy reflect.Value, t reflect.Type) bool {
+	// If there were no FilterValues, we will not detect invalid inputs,
+	// so manually check for them and append invalid if necessary.
+	// We still evaluate the options since an ignore can override invalid.
+	opts := s.opts
+	if !vx.IsValid() || !vy.IsValid() {
+		opts = Options{opts, invalid{}}
+	}
+
+	// Evaluate all filters and apply the remaining options.
+	if opt := opts.filter(s, vx, vy, t); opt != nil {
+		return opt.apply(s, vx, vy)
+	}
+	return false
+}
+
+func (s *state) tryMethod(vx, vy reflect.Value, t reflect.Type) bool {
+	// Check if this type even has an Equal method.
+	m, ok := t.MethodByName("Equal")
+	if !ok || !function.IsType(m.Type, function.EqualAssignable) {
+		return false
+	}
+
+	eq := s.callTTBFunc(m.Func, vx, vy)
+	s.report(eq, vx, vy)
+	return true
+}
+
+func (s *state) callTRFunc(f, v reflect.Value) reflect.Value {
+	if !s.dynChecker.Next() {
+		return f.Call([]reflect.Value{v})[0]
+	}
+
+	// Run the function twice and ensure that we get the same results back.
+	// We run in goroutines so that the race detector (if enabled) can detect
+	// unsafe mutations to the input.
+	c := make(chan reflect.Value)
+	go detectRaces(c, f, v)
+	want := f.Call([]reflect.Value{v})[0]
+	if got := <-c; !s.statelessCompare(got, want).Equal() {
+		// To avoid false-positives with non-reflexive equality operations,
+		// we sanity check whether a value is equal to itself.
+		if !s.statelessCompare(want, want).Equal() {
+			return want
+		}
+		fn := getFuncName(f.Pointer())
+		panic(fmt.Sprintf("non-deterministic function detected: %s", fn))
+	}
+	return want
+}
+
+func (s *state) callTTBFunc(f, x, y reflect.Value) bool {
+	if !s.dynChecker.Next() {
+		return f.Call([]reflect.Value{x, y})[0].Bool()
+	}
+
+	// Swapping the input arguments is sufficient to check that
+	// f is symmetric and deterministic.
+	// We run in goroutines so that the race detector (if enabled) can detect
+	// unsafe mutations to the input.
+	c := make(chan reflect.Value)
+	go detectRaces(c, f, y, x)
+	want := f.Call([]reflect.Value{x, y})[0].Bool()
+	if got := <-c; !got.IsValid() || got.Bool() != want {
+		fn := getFuncName(f.Pointer())
+		panic(fmt.Sprintf("non-deterministic or non-symmetric function detected: %s", fn))
+	}
+	return want
+}
+
+func detectRaces(c chan<- reflect.Value, f reflect.Value, vs ...reflect.Value) {
+	var ret reflect.Value
+	defer func() {
+		recover() // Ignore panics, let the other call to f panic instead
+		c <- ret
+	}()
+	ret = f.Call(vs)[0]
+}
+
+func (s *state) compareArray(vx, vy reflect.Value, t reflect.Type) {
+	step := &sliceIndex{pathStep{t.Elem()}, 0, 0}
+	s.curPath.push(step)
+
+	// Compute an edit-script for slices vx and vy.
+	eq, es := diff.Difference(vx.Len(), vy.Len(), func(ix, iy int) diff.Result {
+		step.xkey, step.ykey = ix, iy
+		return s.statelessCompare(vx.Index(ix), vy.Index(iy))
+	})
+
+	// Equal or no edit-script, so report entire slices as is.
+	if eq || es == nil {
+		s.curPath.pop() // Pop first since we are reporting the whole slice
+		s.report(eq, vx, vy)
+		return
+	}
+
+	// Replay the edit-script.
+	var ix, iy int
+	for _, e := range es {
+		switch e {
+		case diff.UniqueX:
+			step.xkey, step.ykey = ix, -1
+			s.report(false, vx.Index(ix), nothing)
+			ix++
+		case diff.UniqueY:
+			step.xkey, step.ykey = -1, iy
+			s.report(false, nothing, vy.Index(iy))
+			iy++
+		default:
+			step.xkey, step.ykey = ix, iy
+			if e == diff.Identity {
+				s.report(true, vx.Index(ix), vy.Index(iy))
+			} else {
+				s.compareAny(vx.Index(ix), vy.Index(iy))
+			}
+			ix++
+			iy++
+		}
+	}
+	s.curPath.pop()
+	return
+}
+
+func (s *state) compareMap(vx, vy reflect.Value, t reflect.Type) {
+	if vx.IsNil() || vy.IsNil() {
+		s.report(vx.IsNil() && vy.IsNil(), vx, vy)
+		return
+	}
+
+	// We combine and sort the two map keys so that we can perform the
+	// comparisons in a deterministic order.
+	step := &mapIndex{pathStep: pathStep{t.Elem()}}
+	s.curPath.push(step)
+	defer s.curPath.pop()
+	for _, k := range value.SortKeys(append(vx.MapKeys(), vy.MapKeys()...)) {
+		step.key = k
+		vvx := vx.MapIndex(k)
+		vvy := vy.MapIndex(k)
+		switch {
+		case vvx.IsValid() && vvy.IsValid():
+			s.compareAny(vvx, vvy)
+		case vvx.IsValid() && !vvy.IsValid():
+			s.report(false, vvx, nothing)
+		case !vvx.IsValid() && vvy.IsValid():
+			s.report(false, nothing, vvy)
+		default:
+			// It is possible for both vvx and vvy to be invalid if the
+			// key contained a NaN value in it. There is no way in
+			// reflection to be able to retrieve these values.
+			// See https://golang.org/issue/11104
+			panic(fmt.Sprintf("%#v has map key with NaNs", s.curPath))
+		}
+	}
+}
+
+func (s *state) compareStruct(vx, vy reflect.Value, t reflect.Type) {
+	var vax, vay reflect.Value // Addressable versions of vx and vy
+
+	step := &structField{}
+	s.curPath.push(step)
+	defer s.curPath.pop()
+	for i := 0; i < t.NumField(); i++ {
+		vvx := vx.Field(i)
+		vvy := vy.Field(i)
+		step.typ = t.Field(i).Type
+		step.name = t.Field(i).Name
+		step.idx = i
+		step.unexported = !isExported(step.name)
+		if step.unexported {
+			// Defer checking of unexported fields until later to give an
+			// Ignore a chance to ignore the field.
+			if !vax.IsValid() || !vay.IsValid() {
+				// For unsafeRetrieveField to work, the parent struct must
+				// be addressable. Create a new copy of the values if
+				// necessary to make them addressable.
+				vax = makeAddressable(vx)
+				vay = makeAddressable(vy)
+			}
+			step.force = s.exporters[t]
+			step.pvx = vax
+			step.pvy = vay
+			step.field = t.Field(i)
+		}
+		s.compareAny(vvx, vvy)
+	}
+}
+
+// report records the result of a single comparison.
+// It also calls Report if any reporter is registered.
+func (s *state) report(eq bool, vx, vy reflect.Value) {
+	if eq {
+		s.result.NSame++
+	} else {
+		s.result.NDiff++
+	}
+	if s.reporter != nil {
+		s.reporter.Report(vx, vy, eq, s.curPath)
+	}
+}
+
+// dynChecker tracks the state needed to periodically perform checks that
+// user provided functions are symmetric and deterministic.
+// The zero value is safe for immediate use.
+type dynChecker struct{ curr, next int }
+
+// Next increments the state and reports whether a check should be performed.
+//
+// Checks occur every Nth function call, where N is a triangular number:
+//	0 1 3 6 10 15 21 28 36 45 55 66 78 91 105 120 136 153 171 190 ...
+// See https://en.wikipedia.org/wiki/Triangular_number
+//
+// This sequence ensures that the cost of checks drops significantly as
+// the number of functions calls grows larger.
+func (dc *dynChecker) Next() bool {
+	ok := dc.curr == dc.next
+	if ok {
+		dc.curr = 0
+		dc.next++
+	}
+	dc.curr++
+	return ok
+}
+
+// makeAddressable returns a value that is always addressable.
+// It returns the input verbatim if it is already addressable,
+// otherwise it creates a new value and returns an addressable copy.
+func makeAddressable(v reflect.Value) reflect.Value {
+	if v.CanAddr() {
+		return v
+	}
+	vc := reflect.New(v.Type()).Elem()
+	vc.Set(v)
+	return vc
+}
diff --git a/vendor/github.com/google/go-cmp/cmp/internal/diff/debug_disable.go b/vendor/github.com/google/go-cmp/cmp/internal/diff/debug_disable.go
new file mode 100644
index 000000000..42afa4960
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/internal/diff/debug_disable.go
@@ -0,0 +1,17 @@
+// Copyright 2017, 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.md file.
+
+// +build !debug
+
+package diff
+
+var debug debugger
+
+type debugger struct{}
+
+func (debugger) Begin(_, _ int, f EqualFunc, _, _ *EditScript) EqualFunc {
+	return f
+}
+func (debugger) Update() {}
+func (debugger) Finish() {}
diff --git a/vendor/github.com/google/go-cmp/cmp/internal/diff/debug_enable.go b/vendor/github.com/google/go-cmp/cmp/internal/diff/debug_enable.go
new file mode 100644
index 000000000..ba46c62e1
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/internal/diff/debug_enable.go
@@ -0,0 +1,122 @@
+// Copyright 2017, 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.md file.
+
+// +build debug
+
+package diff
+
+import (
+	"fmt"
+	"strings"
+	"sync"
+	"time"
+)
+
+// The algorithm can be seen running in real-time by enabling debugging:
+//	go test -tags=debug -v
+//
+// Example output:
+//	=== RUN   TestDifference/#34
+//	┌───────────────────────────────┐
+//	│ \ · · · · · · · · · · · · · · │
+//	│ · # · · · · · · · · · · · · · │
+//	│ · \ · · · · · · · · · · · · · │
+//	│ · · \ · · · · · · · · · · · · │
+//	│ · · · X # · · · · · · · · · · │
+//	│ · · · # \ · · · · · · · · · · │
+//	│ · · · · · # # · · · · · · · · │
+//	│ · · · · · # \ · · · · · · · · │
+//	│ · · · · · · · \ · · · · · · · │
+//	│ · · · · · · · · \ · · · · · · │
+//	│ · · · · · · · · · \ · · · · · │
+//	│ · · · · · · · · · · \ · · # · │
+//	│ · · · · · · · · · · · \ # # · │
+//	│ · · · · · · · · · · · # # # · │
+//	│ · · · · · · · · · · # # # # · │
+//	│ · · · · · · · · · # # # # # · │
+//	│ · · · · · · · · · · · · · · \ │
+//	└───────────────────────────────┘
+//	[.Y..M.XY......YXYXY.|]
+//
+// The grid represents the edit-graph where the horizontal axis represents
+// list X and the vertical axis represents list Y. The start of the two lists
+// is the top-left, while the ends are the bottom-right. The '·' represents
+// an unexplored node in the graph. The '\' indicates that the two symbols
+// from list X and Y are equal. The 'X' indicates that two symbols are similar
+// (but not exactly equal) to each other. The '#' indicates that the two symbols
+// are different (and not similar). The algorithm traverses this graph trying to
+// make the paths starting in the top-left and the bottom-right connect.
+//
+// The series of '.', 'X', 'Y', and 'M' characters at the bottom represents
+// the currently established path from the forward and reverse searches,
+// seperated by a '|' character.
+
+const (
+	updateDelay  = 100 * time.Millisecond
+	finishDelay  = 500 * time.Millisecond
+	ansiTerminal = true // ANSI escape codes used to move terminal cursor
+)
+
+var debug debugger
+
+type debugger struct {
+	sync.Mutex
+	p1, p2           EditScript
+	fwdPath, revPath *EditScript
+	grid             []byte
+	lines            int
+}
+
+func (dbg *debugger) Begin(nx, ny int, f EqualFunc, p1, p2 *EditScript) EqualFunc {
+	dbg.Lock()
+	dbg.fwdPath, dbg.revPath = p1, p2
+	top := "┌─" + strings.Repeat("──", nx) + "┐\n"
+	row := "│ " + strings.Repeat("· ", nx) + "│\n"
+	btm := "└─" + strings.Repeat("──", nx) + "┘\n"
+	dbg.grid = []byte(top + strings.Repeat(row, ny) + btm)
+	dbg.lines = strings.Count(dbg.String(), "\n")
+	fmt.Print(dbg)
+
+	// Wrap the EqualFunc so that we can intercept each result.
+	return func(ix, iy int) (r Result) {
+		cell := dbg.grid[len(top)+iy*len(row):][len("│ ")+len("· ")*ix:][:len("·")]
+		for i := range cell {
+			cell[i] = 0 // Zero out the multiple bytes of UTF-8 middle-dot
+		}
+		switch r = f(ix, iy); {
+		case r.Equal():
+			cell[0] = '\\'
+		case r.Similar():
+			cell[0] = 'X'
+		default:
+			cell[0] = '#'
+		}
+		return
+	}
+}
+
+func (dbg *debugger) Update() {
+	dbg.print(updateDelay)
+}
+
+func (dbg *debugger) Finish() {
+	dbg.print(finishDelay)
+	dbg.Unlock()
+}
+
+func (dbg *debugger) String() string {
+	dbg.p1, dbg.p2 = *dbg.fwdPath, dbg.p2[:0]
+	for i := len(*dbg.revPath) - 1; i >= 0; i-- {
+		dbg.p2 = append(dbg.p2, (*dbg.revPath)[i])
+	}
+	return fmt.Sprintf("%s[%v|%v]\n\n", dbg.grid, dbg.p1, dbg.p2)
+}
+
+func (dbg *debugger) print(d time.Duration) {
+	if ansiTerminal {
+		fmt.Printf("\x1b[%dA", dbg.lines) // Reset terminal cursor
+	}
+	fmt.Print(dbg)
+	time.Sleep(d)
+}
diff --git a/vendor/github.com/google/go-cmp/cmp/internal/diff/diff.go b/vendor/github.com/google/go-cmp/cmp/internal/diff/diff.go
new file mode 100644
index 000000000..baa41fd23
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/internal/diff/diff.go
@@ -0,0 +1,373 @@
+// Copyright 2017, 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.md file.
+
+// Package diff implements an algorithm for producing edit-scripts.
+// The edit-script is a sequence of operations needed to transform one list
+// of symbols into another (or vice-versa). The edits allowed are insertions,
+// deletions, and modifications. The summation of all edits is called the
+// Levenshtein distance as this problem is well-known in computer science.
+//
+// This package prioritizes performance over accuracy. That is, the run time
+// is more important than obtaining a minimal Levenshtein distance.
+package diff
+
+// EditType represents a single operation within an edit-script.
+type EditType uint8
+
+const (
+	// Identity indicates that a symbol pair is identical in both list X and Y.
+	Identity EditType = iota
+	// UniqueX indicates that a symbol only exists in X and not Y.
+	UniqueX
+	// UniqueY indicates that a symbol only exists in Y and not X.
+	UniqueY
+	// Modified indicates that a symbol pair is a modification of each other.
+	Modified
+)
+
+// EditScript represents the series of differences between two lists.
+type EditScript []EditType
+
+// String returns a human-readable string representing the edit-script where
+// Identity, UniqueX, UniqueY, and Modified are represented by the
+// '.', 'X', 'Y', and 'M' characters, respectively.
+func (es EditScript) String() string {
+	b := make([]byte, len(es))
+	for i, e := range es {
+		switch e {
+		case Identity:
+			b[i] = '.'
+		case UniqueX:
+			b[i] = 'X'
+		case UniqueY:
+			b[i] = 'Y'
+		case Modified:
+			b[i] = 'M'
+		default:
+			panic("invalid edit-type")
+		}
+	}
+	return string(b)
+}
+
+// stats returns a histogram of the number of each type of edit operation.
+func (es EditScript) stats() (s struct{ NI, NX, NY, NM int }) {
+	for _, e := range es {
+		switch e {
+		case Identity:
+			s.NI++
+		case UniqueX:
+			s.NX++
+		case UniqueY:
+			s.NY++
+		case Modified:
+			s.NM++
+		default:
+			panic("invalid edit-type")
+		}
+	}
+	return
+}
+
+// Dist is the Levenshtein distance and is guaranteed to be 0 if and only if
+// lists X and Y are equal.
+func (es EditScript) Dist() int { return len(es) - es.stats().NI }
+
+// LenX is the length of the X list.
+func (es EditScript) LenX() int { return len(es) - es.stats().NY }
+
+// LenY is the length of the Y list.
+func (es EditScript) LenY() int { return len(es) - es.stats().NX }
+
+// EqualFunc reports whether the symbols at indexes ix and iy are equal.
+// When called by Difference, the index is guaranteed to be within nx and ny.
+type EqualFunc func(ix int, iy int) Result
+
+// Result is the result of comparison.
+// NSame is the number of sub-elements that are equal.
+// NDiff is the number of sub-elements that are not equal.
+type Result struct{ NSame, NDiff int }
+
+// Equal indicates whether the symbols are equal. Two symbols are equal
+// if and only if NDiff == 0. If Equal, then they are also Similar.
+func (r Result) Equal() bool { return r.NDiff == 0 }
+
+// Similar indicates whether two symbols are similar and may be represented
+// by using the Modified type. As a special case, we consider binary comparisons
+// (i.e., those that return Result{1, 0} or Result{0, 1}) to be similar.
+//
+// The exact ratio of NSame to NDiff to determine similarity may change.
+func (r Result) Similar() bool {
+	// Use NSame+1 to offset NSame so that binary comparisons are similar.
+	return r.NSame+1 >= r.NDiff
+}
+
+// Difference reports whether two lists of lengths nx and ny are equal
+// given the definition of equality provided as f.
+//
+// This function may return a edit-script, which is a sequence of operations
+// needed to convert one list into the other. If non-nil, the following
+// invariants for the edit-script are maintained:
+//	• eq == (es.Dist()==0)
+//	• nx == es.LenX()
+//	• ny == es.LenY()
+//
+// This algorithm is not guaranteed to be an optimal solution (i.e., one that
+// produces an edit-script with a minimal Levenshtein distance). This algorithm
+// favors performance over optimality. The exact output is not guaranteed to
+// be stable and may change over time.
+func Difference(nx, ny int, f EqualFunc) (eq bool, es EditScript) {
+	es = searchGraph(nx, ny, f)
+	st := es.stats()
+	eq = len(es) == st.NI
+	if !eq && st.NI < (nx+ny)/4 {
+		return eq, nil // Edit-script more distracting than helpful
+	}
+	return eq, es
+}
+
+func searchGraph(nx, ny int, f EqualFunc) EditScript {
+	// This algorithm is based on traversing what is known as an "edit-graph".
+	// See Figure 1 from "An O(ND) Difference Algorithm and Its Variations"
+	// by Eugene W. Myers. Since D can be as large as N itself, this is
+	// effectively O(N^2). Unlike the algorithm from that paper, we are not
+	// interested in the optimal path, but at least some "decent" path.
+	//
+	// For example, let X and Y be lists of symbols:
+	//	X = [A B C A B B A]
+	//	Y = [C B A B A C]
+	//
+	// The edit-graph can be drawn as the following:
+	//	   A B C A B B A
+	//	  ┌─────────────┐
+	//	C │_|_|\|_|_|_|_│ 0
+	//	B │_|\|_|_|\|\|_│ 1
+	//	A │\|_|_|\|_|_|\│ 2
+	//	B │_|\|_|_|\|\|_│ 3
+	//	A │\|_|_|\|_|_|\│ 4
+	//	C │ | |\| | | | │ 5
+	//	  └─────────────┘ 6
+	//	   0 1 2 3 4 5 6 7
+	//
+	// List X is written along the horizontal axis, while list Y is written
+	// along the vertical axis. At any point on this grid, if the symbol in
+	// list X matches the corresponding symbol in list Y, then a '\' is drawn.
+	// The goal of any minimal edit-script algorithm is to find a path from the
+	// top-left corner to the bottom-right corner, while traveling through the
+	// fewest horizontal or vertical edges.
+	// A horizontal edge is equivalent to inserting a symbol from list X.
+	// A vertical edge is equivalent to inserting a symbol from list Y.
+	// A diagonal edge is equivalent to a matching symbol between both X and Y.
+
+	// Invariants:
+	//	• 0 ≤ fwdPath.X ≤ (fwdFrontier.X, revFrontier.X) ≤ revPath.X ≤ nx
+	//	• 0 ≤ fwdPath.Y ≤ (fwdFrontier.Y, revFrontier.Y) ≤ revPath.Y ≤ ny
+	//
+	// In general:
+	//	• fwdFrontier.X < revFrontier.X
+	//	• fwdFrontier.Y < revFrontier.Y
+	// Unless, it is time for the algorithm to terminate.
+	fwdPath := path{+1, point{0, 0}, make(EditScript, 0, (nx+ny)/2)}
+	revPath := path{-1, point{nx, ny}, make(EditScript, 0)}
+	fwdFrontier := fwdPath.point // Forward search frontier
+	revFrontier := revPath.point // Reverse search frontier
+
+	// Search budget bounds the cost of searching for better paths.
+	// The longest sequence of non-matching symbols that can be tolerated is
+	// approximately the square-root of the search budget.
+	searchBudget := 4 * (nx + ny) // O(n)
+
+	// The algorithm below is a greedy, meet-in-the-middle algorithm for
+	// computing sub-optimal edit-scripts between two lists.
+	//
+	// The algorithm is approximately as follows:
+	//	• Searching for differences switches back-and-forth between
+	//	a search that starts at the beginning (the top-left corner), and
+	//	a search that starts at the end (the bottom-right corner). The goal of
+	//	the search is connect with the search from the opposite corner.
+	//	• As we search, we build a path in a greedy manner, where the first
+	//	match seen is added to the path (this is sub-optimal, but provides a
+	//	decent result in practice). When matches are found, we try the next pair
+	//	of symbols in the lists and follow all matches as far as possible.
+	//	• When searching for matches, we search along a diagonal going through
+	//	through the "frontier" point. If no matches are found, we advance the
+	//	frontier towards the opposite corner.
+	//	• This algorithm terminates when either the X coordinates or the
+	//	Y coordinates of the forward and reverse frontier points ever intersect.
+	//
+	// This algorithm is correct even if searching only in the forward direction
+	// or in the reverse direction. We do both because it is commonly observed
+	// that two lists commonly differ because elements were added to the front
+	// or end of the other list.
+	//
+	// Running the tests with the "debug" build tag prints a visualization of
+	// the algorithm running in real-time. This is educational for understanding
+	// how the algorithm works. See debug_enable.go.
+	f = debug.Begin(nx, ny, f, &fwdPath.es, &revPath.es)
+	for {
+		// Forward search from the beginning.
+		if fwdFrontier.X >= revFrontier.X || fwdFrontier.Y >= revFrontier.Y || searchBudget == 0 {
+			break
+		}
+		for stop1, stop2, i := false, false, 0; !(stop1 && stop2) && searchBudget > 0; i++ {
+			// Search in a diagonal pattern for a match.
+			z := zigzag(i)
+			p := point{fwdFrontier.X + z, fwdFrontier.Y - z}
+			switch {
+			case p.X >= revPath.X || p.Y < fwdPath.Y:
+				stop1 = true // Hit top-right corner
+			case p.Y >= revPath.Y || p.X < fwdPath.X:
+				stop2 = true // Hit bottom-left corner
+			case f(p.X, p.Y).Equal():
+				// Match found, so connect the path to this point.
+				fwdPath.connect(p, f)
+				fwdPath.append(Identity)
+				// Follow sequence of matches as far as possible.
+				for fwdPath.X < revPath.X && fwdPath.Y < revPath.Y {
+					if !f(fwdPath.X, fwdPath.Y).Equal() {
+						break
+					}
+					fwdPath.append(Identity)
+				}
+				fwdFrontier = fwdPath.point
+				stop1, stop2 = true, true
+			default:
+				searchBudget-- // Match not found
+			}
+			debug.Update()
+		}
+		// Advance the frontier towards reverse point.
+		if revPath.X-fwdFrontier.X >= revPath.Y-fwdFrontier.Y {
+			fwdFrontier.X++
+		} else {
+			fwdFrontier.Y++
+		}
+
+		// Reverse search from the end.
+		if fwdFrontier.X >= revFrontier.X || fwdFrontier.Y >= revFrontier.Y || searchBudget == 0 {
+			break
+		}
+		for stop1, stop2, i := false, false, 0; !(stop1 && stop2) && searchBudget > 0; i++ {
+			// Search in a diagonal pattern for a match.
+			z := zigzag(i)
+			p := point{revFrontier.X - z, revFrontier.Y + z}
+			switch {
+			case fwdPath.X >= p.X || revPath.Y < p.Y:
+				stop1 = true // Hit bottom-left corner
+			case fwdPath.Y >= p.Y || revPath.X < p.X:
+				stop2 = true // Hit top-right corner
+			case f(p.X-1, p.Y-1).Equal():
+				// Match found, so connect the path to this point.
+				revPath.connect(p, f)
+				revPath.append(Identity)
+				// Follow sequence of matches as far as possible.
+				for fwdPath.X < revPath.X && fwdPath.Y < revPath.Y {
+					if !f(revPath.X-1, revPath.Y-1).Equal() {
+						break
+					}
+					revPath.append(Identity)
+				}
+				revFrontier = revPath.point
+				stop1, stop2 = true, true
+			default:
+				searchBudget-- // Match not found
+			}
+			debug.Update()
+		}
+		// Advance the frontier towards forward point.
+		if revFrontier.X-fwdPath.X >= revFrontier.Y-fwdPath.Y {
+			revFrontier.X--
+		} else {
+			revFrontier.Y--
+		}
+	}
+
+	// Join the forward and reverse paths and then append the reverse path.
+	fwdPath.connect(revPath.point, f)
+	for i := len(revPath.es) - 1; i >= 0; i-- {
+		t := revPath.es[i]
+		revPath.es = revPath.es[:i]
+		fwdPath.append(t)
+	}
+	debug.Finish()
+	return fwdPath.es
+}
+
+type path struct {
+	dir   int // +1 if forward, -1 if reverse
+	point     // Leading point of the EditScript path
+	es    EditScript
+}
+
+// connect appends any necessary Identity, Modified, UniqueX, or UniqueY types
+// to the edit-script to connect p.point to dst.
+func (p *path) connect(dst point, f EqualFunc) {
+	if p.dir > 0 {
+		// Connect in forward direction.
+		for dst.X > p.X && dst.Y > p.Y {
+			switch r := f(p.X, p.Y); {
+			case r.Equal():
+				p.append(Identity)
+			case r.Similar():
+				p.append(Modified)
+			case dst.X-p.X >= dst.Y-p.Y:
+				p.append(UniqueX)
+			default:
+				p.append(UniqueY)
+			}
+		}
+		for dst.X > p.X {
+			p.append(UniqueX)
+		}
+		for dst.Y > p.Y {
+			p.append(UniqueY)
+		}
+	} else {
+		// Connect in reverse direction.
+		for p.X > dst.X && p.Y > dst.Y {
+			switch r := f(p.X-1, p.Y-1); {
+			case r.Equal():
+				p.append(Identity)
+			case r.Similar():
+				p.append(Modified)
+			case p.Y-dst.Y >= p.X-dst.X:
+				p.append(UniqueY)
+			default:
+				p.append(UniqueX)
+			}
+		}
+		for p.X > dst.X {
+			p.append(UniqueX)
+		}
+		for p.Y > dst.Y {
+			p.append(UniqueY)
+		}
+	}
+}
+
+func (p *path) append(t EditType) {
+	p.es = append(p.es, t)
+	switch t {
+	case Identity, Modified:
+		p.add(p.dir, p.dir)
+	case UniqueX:
+		p.add(p.dir, 0)
+	case UniqueY:
+		p.add(0, p.dir)
+	}
+	debug.Update()
+}
+
+type point struct{ X, Y int }
+
+func (p *point) add(dx, dy int) { p.X += dx; p.Y += dy }
+
+// zigzag maps a consecutive sequence of integers to a zig-zag sequence.
+//	[0 1 2 3 4 5 ...] => [0 -1 +1 -2 +2 ...]
+func zigzag(x int) int {
+	if x&1 != 0 {
+		x = ^x
+	}
+	return x >> 1
+}
diff --git a/vendor/github.com/google/go-cmp/cmp/internal/function/func.go b/vendor/github.com/google/go-cmp/cmp/internal/function/func.go
new file mode 100644
index 000000000..4c35ff11e
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/internal/function/func.go
@@ -0,0 +1,49 @@
+// Copyright 2017, 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.md file.
+
+// Package function identifies function types.
+package function
+
+import "reflect"
+
+type funcType int
+
+const (
+	_ funcType = iota
+
+	ttbFunc // func(T, T) bool
+	tibFunc // func(T, I) bool
+	trFunc  // func(T) R
+
+	Equal           = ttbFunc // func(T, T) bool
+	EqualAssignable = tibFunc // func(T, I) bool; encapsulates func(T, T) bool
+	Transformer     = trFunc  // func(T) R
+	ValueFilter     = ttbFunc // func(T, T) bool
+	Less            = ttbFunc // func(T, T) bool
+)
+
+var boolType = reflect.TypeOf(true)
+
+// IsType reports whether the reflect.Type is of the specified function type.
+func IsType(t reflect.Type, ft funcType) bool {
+	if t == nil || t.Kind() != reflect.Func || t.IsVariadic() {
+		return false
+	}
+	ni, no := t.NumIn(), t.NumOut()
+	switch ft {
+	case ttbFunc: // func(T, T) bool
+		if ni == 2 && no == 1 && t.In(0) == t.In(1) && t.Out(0) == boolType {
+			return true
+		}
+	case tibFunc: // func(T, I) bool
+		if ni == 2 && no == 1 && t.In(0).AssignableTo(t.In(1)) && t.Out(0) == boolType {
+			return true
+		}
+	case trFunc: // func(T) R
+		if ni == 1 && no == 1 {
+			return true
+		}
+	}
+	return false
+}
diff --git a/vendor/github.com/google/go-cmp/cmp/internal/value/format.go b/vendor/github.com/google/go-cmp/cmp/internal/value/format.go
new file mode 100644
index 000000000..abaeca89e
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/internal/value/format.go
@@ -0,0 +1,259 @@
+// Copyright 2017, 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.md file.
+
+// Package value provides functionality for reflect.Value types.
+package value
+
+import (
+	"fmt"
+	"reflect"
+	"strings"
+	"unicode"
+	"unicode/utf8"
+)
+
+// formatFakePointers controls whether to substitute pointer addresses with nil.
+// This is used for deterministic testing.
+var formatFakePointers = false
+
+var stringerIface = reflect.TypeOf((*fmt.Stringer)(nil)).Elem()
+
+// Format formats the value v as a string.
+//
+// This is similar to fmt.Sprintf("%+v", v) except this:
+//	* Prints the type unless it can be elided
+//	* Avoids printing struct fields that are zero
+//	* Prints a nil-slice as being nil, not empty
+//	* Prints map entries in deterministic order
+func Format(v reflect.Value, useStringer bool) string {
+	return formatAny(v, formatConfig{useStringer, true, true, !formatFakePointers}, nil)
+}
+
+type formatConfig struct {
+	useStringer    bool // Should the String method be used if available?
+	printType      bool // Should we print the type before the value?
+	followPointers bool // Should we recursively follow pointers?
+	realPointers   bool // Should we print the real address of pointers?
+}
+
+func formatAny(v reflect.Value, conf formatConfig, visited map[uintptr]bool) string {
+	// TODO: Should this be a multi-line printout in certain situations?
+
+	if !v.IsValid() {
+		return "<non-existent>"
+	}
+	if conf.useStringer && v.Type().Implements(stringerIface) && v.CanInterface() {
+		if (v.Kind() == reflect.Ptr || v.Kind() == reflect.Interface) && v.IsNil() {
+			return "<nil>"
+		}
+		return fmt.Sprintf("%q", v.Interface().(fmt.Stringer).String())
+	}
+
+	switch v.Kind() {
+	case reflect.Bool:
+		return formatPrimitive(v.Type(), v.Bool(), conf)
+	case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64:
+		return formatPrimitive(v.Type(), v.Int(), conf)
+	case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr:
+		if v.Type().PkgPath() == "" || v.Kind() == reflect.Uintptr {
+			// Unnamed uints are usually bytes or words, so use hexadecimal.
+			return formatPrimitive(v.Type(), formatHex(v.Uint()), conf)
+		}
+		return formatPrimitive(v.Type(), v.Uint(), conf)
+	case reflect.Float32, reflect.Float64:
+		return formatPrimitive(v.Type(), v.Float(), conf)
+	case reflect.Complex64, reflect.Complex128:
+		return formatPrimitive(v.Type(), v.Complex(), conf)
+	case reflect.String:
+		return formatPrimitive(v.Type(), fmt.Sprintf("%q", v), conf)
+	case reflect.UnsafePointer, reflect.Chan, reflect.Func:
+		return formatPointer(v, conf)
+	case reflect.Ptr:
+		if v.IsNil() {
+			if conf.printType {
+				return fmt.Sprintf("(%v)(nil)", v.Type())
+			}
+			return "<nil>"
+		}
+		if visited[v.Pointer()] || !conf.followPointers {
+			return formatPointer(v, conf)
+		}
+		visited = insertPointer(visited, v.Pointer())
+		return "&" + formatAny(v.Elem(), conf, visited)
+	case reflect.Interface:
+		if v.IsNil() {
+			if conf.printType {
+				return fmt.Sprintf("%v(nil)", v.Type())
+			}
+			return "<nil>"
+		}
+		return formatAny(v.Elem(), conf, visited)
+	case reflect.Slice:
+		if v.IsNil() {
+			if conf.printType {
+				return fmt.Sprintf("%v(nil)", v.Type())
+			}
+			return "<nil>"
+		}
+		if visited[v.Pointer()] {
+			return formatPointer(v, conf)
+		}
+		visited = insertPointer(visited, v.Pointer())
+		fallthrough
+	case reflect.Array:
+		var ss []string
+		subConf := conf
+		subConf.printType = v.Type().Elem().Kind() == reflect.Interface
+		for i := 0; i < v.Len(); i++ {
+			s := formatAny(v.Index(i), subConf, visited)
+			ss = append(ss, s)
+		}
+		s := fmt.Sprintf("{%s}", strings.Join(ss, ", "))
+		if conf.printType {
+			return v.Type().String() + s
+		}
+		return s
+	case reflect.Map:
+		if v.IsNil() {
+			if conf.printType {
+				return fmt.Sprintf("%v(nil)", v.Type())
+			}
+			return "<nil>"
+		}
+		if visited[v.Pointer()] {
+			return formatPointer(v, conf)
+		}
+		visited = insertPointer(visited, v.Pointer())
+
+		var ss []string
+		subConf := conf
+		subConf.printType = v.Type().Elem().Kind() == reflect.Interface
+		for _, k := range SortKeys(v.MapKeys()) {
+			sk := formatAny(k, formatConfig{realPointers: conf.realPointers}, visited)
+			sv := formatAny(v.MapIndex(k), subConf, visited)
+			ss = append(ss, fmt.Sprintf("%s: %s", sk, sv))
+		}
+		s := fmt.Sprintf("{%s}", strings.Join(ss, ", "))
+		if conf.printType {
+			return v.Type().String() + s
+		}
+		return s
+	case reflect.Struct:
+		var ss []string
+		subConf := conf
+		subConf.printType = true
+		for i := 0; i < v.NumField(); i++ {
+			vv := v.Field(i)
+			if isZero(vv) {
+				continue // Elide zero value fields
+			}
+			name := v.Type().Field(i).Name
+			subConf.useStringer = conf.useStringer && isExported(name)
+			s := formatAny(vv, subConf, visited)
+			ss = append(ss, fmt.Sprintf("%s: %s", name, s))
+		}
+		s := fmt.Sprintf("{%s}", strings.Join(ss, ", "))
+		if conf.printType {
+			return v.Type().String() + s
+		}
+		return s
+	default:
+		panic(fmt.Sprintf("%v kind not handled", v.Kind()))
+	}
+}
+
+func formatPrimitive(t reflect.Type, v interface{}, conf formatConfig) string {
+	if conf.printType && t.PkgPath() != "" {
+		return fmt.Sprintf("%v(%v)", t, v)
+	}
+	return fmt.Sprintf("%v", v)
+}
+
+func formatPointer(v reflect.Value, conf formatConfig) string {
+	p := v.Pointer()
+	if !conf.realPointers {
+		p = 0 // For deterministic printing purposes
+	}
+	s := formatHex(uint64(p))
+	if conf.printType {
+		return fmt.Sprintf("(%v)(%s)", v.Type(), s)
+	}
+	return s
+}
+
+func formatHex(u uint64) string {
+	var f string
+	switch {
+	case u <= 0xff:
+		f = "0x%02x"
+	case u <= 0xffff:
+		f = "0x%04x"
+	case u <= 0xffffff:
+		f = "0x%06x"
+	case u <= 0xffffffff:
+		f = "0x%08x"
+	case u <= 0xffffffffff:
+		f = "0x%010x"
+	case u <= 0xffffffffffff:
+		f = "0x%012x"
+	case u <= 0xffffffffffffff:
+		f = "0x%014x"
+	case u <= 0xffffffffffffffff:
+		f = "0x%016x"
+	}
+	return fmt.Sprintf(f, u)
+}
+
+// insertPointer insert p into m, allocating m if necessary.
+func insertPointer(m map[uintptr]bool, p uintptr) map[uintptr]bool {
+	if m == nil {
+		m = make(map[uintptr]bool)
+	}
+	m[p] = true
+	return m
+}
+
+// isZero reports whether v is the zero value.
+// This does not rely on Interface and so can be used on unexported fields.
+func isZero(v reflect.Value) bool {
+	switch v.Kind() {
+	case reflect.Bool:
+		return v.Bool() == false
+	case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64:
+		return v.Int() == 0
+	case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr:
+		return v.Uint() == 0
+	case reflect.Float32, reflect.Float64:
+		return v.Float() == 0
+	case reflect.Complex64, reflect.Complex128:
+		return v.Complex() == 0
+	case reflect.String:
+		return v.String() == ""
+	case reflect.UnsafePointer:
+		return v.Pointer() == 0
+	case reflect.Chan, reflect.Func, reflect.Interface, reflect.Ptr, reflect.Map, reflect.Slice:
+		return v.IsNil()
+	case reflect.Array:
+		for i := 0; i < v.Len(); i++ {
+			if !isZero(v.Index(i)) {
+				return false
+			}
+		}
+		return true
+	case reflect.Struct:
+		for i := 0; i < v.NumField(); i++ {
+			if !isZero(v.Field(i)) {
+				return false
+			}
+		}
+		return true
+	}
+	return false
+}
+
+// isExported reports whether the identifier is exported.
+func isExported(id string) bool {
+	r, _ := utf8.DecodeRuneInString(id)
+	return unicode.IsUpper(r)
+}
diff --git a/vendor/github.com/google/go-cmp/cmp/internal/value/sort.go b/vendor/github.com/google/go-cmp/cmp/internal/value/sort.go
new file mode 100644
index 000000000..fe8aa27a0
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/internal/value/sort.go
@@ -0,0 +1,111 @@
+// Copyright 2017, 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.md file.
+
+package value
+
+import (
+	"fmt"
+	"math"
+	"reflect"
+	"sort"
+)
+
+// SortKeys sorts a list of map keys, deduplicating keys if necessary.
+// The type of each value must be comparable.
+func SortKeys(vs []reflect.Value) []reflect.Value {
+	if len(vs) == 0 {
+		return vs
+	}
+
+	// Sort the map keys.
+	sort.Sort(valueSorter(vs))
+
+	// Deduplicate keys (fails for NaNs).
+	vs2 := vs[:1]
+	for _, v := range vs[1:] {
+		if isLess(vs2[len(vs2)-1], v) {
+			vs2 = append(vs2, v)
+		}
+	}
+	return vs2
+}
+
+// TODO: Use sort.Slice once Google AppEngine is on Go1.8 or above.
+type valueSorter []reflect.Value
+
+func (vs valueSorter) Len() int           { return len(vs) }
+func (vs valueSorter) Less(i, j int) bool { return isLess(vs[i], vs[j]) }
+func (vs valueSorter) Swap(i, j int)      { vs[i], vs[j] = vs[j], vs[i] }
+
+// isLess is a generic function for sorting arbitrary map keys.
+// The inputs must be of the same type and must be comparable.
+func isLess(x, y reflect.Value) bool {
+	switch x.Type().Kind() {
+	case reflect.Bool:
+		return !x.Bool() && y.Bool()
+	case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64:
+		return x.Int() < y.Int()
+	case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr:
+		return x.Uint() < y.Uint()
+	case reflect.Float32, reflect.Float64:
+		fx, fy := x.Float(), y.Float()
+		return fx < fy || math.IsNaN(fx) && !math.IsNaN(fy)
+	case reflect.Complex64, reflect.Complex128:
+		cx, cy := x.Complex(), y.Complex()
+		rx, ix, ry, iy := real(cx), imag(cx), real(cy), imag(cy)
+		if rx == ry || (math.IsNaN(rx) && math.IsNaN(ry)) {
+			return ix < iy || math.IsNaN(ix) && !math.IsNaN(iy)
+		}
+		return rx < ry || math.IsNaN(rx) && !math.IsNaN(ry)
+	case reflect.Ptr, reflect.UnsafePointer, reflect.Chan:
+		return x.Pointer() < y.Pointer()
+	case reflect.String:
+		return x.String() < y.String()
+	case reflect.Array:
+		for i := 0; i < x.Len(); i++ {
+			if isLess(x.Index(i), y.Index(i)) {
+				return true
+			}
+			if isLess(y.Index(i), x.Index(i)) {
+				return false
+			}
+		}
+		return false
+	case reflect.Struct:
+		for i := 0; i < x.NumField(); i++ {
+			if isLess(x.Field(i), y.Field(i)) {
+				return true
+			}
+			if isLess(y.Field(i), x.Field(i)) {
+				return false
+			}
+		}
+		return false
+	case reflect.Interface:
+		vx, vy := x.Elem(), y.Elem()
+		if !vx.IsValid() || !vy.IsValid() {
+			return !vx.IsValid() && vy.IsValid()
+		}
+		tx, ty := vx.Type(), vy.Type()
+		if tx == ty {
+			return isLess(x.Elem(), y.Elem())
+		}
+		if tx.Kind() != ty.Kind() {
+			return vx.Kind() < vy.Kind()
+		}
+		if tx.String() != ty.String() {
+			return tx.String() < ty.String()
+		}
+		if tx.PkgPath() != ty.PkgPath() {
+			return tx.PkgPath() < ty.PkgPath()
+		}
+		// This can happen in rare situations, so we fallback to just comparing
+		// the unique pointer for a reflect.Type. This guarantees deterministic
+		// ordering within a program, but it is obviously not stable.
+		return reflect.ValueOf(vx.Type()).Pointer() < reflect.ValueOf(vy.Type()).Pointer()
+	default:
+		// Must be Func, Map, or Slice; which are not comparable.
+		panic(fmt.Sprintf("%T is not comparable", x.Type()))
+	}
+}
diff --git a/vendor/github.com/google/go-cmp/cmp/options.go b/vendor/github.com/google/go-cmp/cmp/options.go
new file mode 100644
index 000000000..a4e159ac4
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/options.go
@@ -0,0 +1,446 @@
+// Copyright 2017, 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.md file.
+
+package cmp
+
+import (
+	"fmt"
+	"reflect"
+	"runtime"
+	"strings"
+
+	"github.com/google/go-cmp/cmp/internal/function"
+)
+
+// Option configures for specific behavior of Equal and Diff. In particular,
+// the fundamental Option functions (Ignore, Transformer, and Comparer),
+// configure how equality is determined.
+//
+// The fundamental options may be composed with filters (FilterPath and
+// FilterValues) to control the scope over which they are applied.
+//
+// The cmp/cmpopts package provides helper functions for creating options that
+// may be used with Equal and Diff.
+type Option interface {
+	// filter applies all filters and returns the option that remains.
+	// Each option may only read s.curPath and call s.callTTBFunc.
+	//
+	// An Options is returned only if multiple comparers or transformers
+	// can apply simultaneously and will only contain values of those types
+	// or sub-Options containing values of those types.
+	filter(s *state, vx, vy reflect.Value, t reflect.Type) applicableOption
+}
+
+// applicableOption represents the following types:
+//	Fundamental: ignore | invalid | *comparer | *transformer
+//	Grouping:    Options
+type applicableOption interface {
+	Option
+
+	// apply executes the option and reports whether the option was applied.
+	// Each option may mutate s.
+	apply(s *state, vx, vy reflect.Value) bool
+}
+
+// coreOption represents the following types:
+//	Fundamental: ignore | invalid | *comparer | *transformer
+//	Filters:     *pathFilter | *valuesFilter
+type coreOption interface {
+	Option
+	isCore()
+}
+
+type core struct{}
+
+func (core) isCore() {}
+
+// Options is a list of Option values that also satisfies the Option interface.
+// Helper comparison packages may return an Options value when packing multiple
+// Option values into a single Option. When this package processes an Options,
+// it will be implicitly expanded into a flat list.
+//
+// Applying a filter on an Options is equivalent to applying that same filter
+// on all individual options held within.
+type Options []Option
+
+func (opts Options) filter(s *state, vx, vy reflect.Value, t reflect.Type) (out applicableOption) {
+	for _, opt := range opts {
+		switch opt := opt.filter(s, vx, vy, t); opt.(type) {
+		case ignore:
+			return ignore{} // Only ignore can short-circuit evaluation
+		case invalid:
+			out = invalid{} // Takes precedence over comparer or transformer
+		case *comparer, *transformer, Options:
+			switch out.(type) {
+			case nil:
+				out = opt
+			case invalid:
+				// Keep invalid
+			case *comparer, *transformer, Options:
+				out = Options{out, opt} // Conflicting comparers or transformers
+			}
+		}
+	}
+	return out
+}
+
+func (opts Options) apply(s *state, _, _ reflect.Value) bool {
+	const warning = "ambiguous set of applicable options"
+	const help = "consider using filters to ensure at most one Comparer or Transformer may apply"
+	var ss []string
+	for _, opt := range flattenOptions(nil, opts) {
+		ss = append(ss, fmt.Sprint(opt))
+	}
+	set := strings.Join(ss, "\n\t")
+	panic(fmt.Sprintf("%s at %#v:\n\t%s\n%s", warning, s.curPath, set, help))
+}
+
+func (opts Options) String() string {
+	var ss []string
+	for _, opt := range opts {
+		ss = append(ss, fmt.Sprint(opt))
+	}
+	return fmt.Sprintf("Options{%s}", strings.Join(ss, ", "))
+}
+
+// FilterPath returns a new Option where opt is only evaluated if filter f
+// returns true for the current Path in the value tree.
+//
+// The option passed in may be an Ignore, Transformer, Comparer, Options, or
+// a previously filtered Option.
+func FilterPath(f func(Path) bool, opt Option) Option {
+	if f == nil {
+		panic("invalid path filter function")
+	}
+	if opt := normalizeOption(opt); opt != nil {
+		return &pathFilter{fnc: f, opt: opt}
+	}
+	return nil
+}
+
+type pathFilter struct {
+	core
+	fnc func(Path) bool
+	opt Option
+}
+
+func (f pathFilter) filter(s *state, vx, vy reflect.Value, t reflect.Type) applicableOption {
+	if f.fnc(s.curPath) {
+		return f.opt.filter(s, vx, vy, t)
+	}
+	return nil
+}
+
+func (f pathFilter) String() string {
+	fn := getFuncName(reflect.ValueOf(f.fnc).Pointer())
+	return fmt.Sprintf("FilterPath(%s, %v)", fn, f.opt)
+}
+
+// FilterValues returns a new Option where opt is only evaluated if filter f,
+// which is a function of the form "func(T, T) bool", returns true for the
+// current pair of values being compared. If the type of the values is not
+// assignable to T, then this filter implicitly returns false.
+//
+// The filter function must be
+// symmetric (i.e., agnostic to the order of the inputs) and
+// deterministic (i.e., produces the same result when given the same inputs).
+// If T is an interface, it is possible that f is called with two values with
+// different concrete types that both implement T.
+//
+// The option passed in may be an Ignore, Transformer, Comparer, Options, or
+// a previously filtered Option.
+func FilterValues(f interface{}, opt Option) Option {
+	v := reflect.ValueOf(f)
+	if !function.IsType(v.Type(), function.ValueFilter) || v.IsNil() {
+		panic(fmt.Sprintf("invalid values filter function: %T", f))
+	}
+	if opt := normalizeOption(opt); opt != nil {
+		vf := &valuesFilter{fnc: v, opt: opt}
+		if ti := v.Type().In(0); ti.Kind() != reflect.Interface || ti.NumMethod() > 0 {
+			vf.typ = ti
+		}
+		return vf
+	}
+	return nil
+}
+
+type valuesFilter struct {
+	core
+	typ reflect.Type  // T
+	fnc reflect.Value // func(T, T) bool
+	opt Option
+}
+
+func (f valuesFilter) filter(s *state, vx, vy reflect.Value, t reflect.Type) applicableOption {
+	if !vx.IsValid() || !vy.IsValid() {
+		return invalid{}
+	}
+	if (f.typ == nil || t.AssignableTo(f.typ)) && s.callTTBFunc(f.fnc, vx, vy) {
+		return f.opt.filter(s, vx, vy, t)
+	}
+	return nil
+}
+
+func (f valuesFilter) String() string {
+	fn := getFuncName(f.fnc.Pointer())
+	return fmt.Sprintf("FilterValues(%s, %v)", fn, f.opt)
+}
+
+// Ignore is an Option that causes all comparisons to be ignored.
+// This value is intended to be combined with FilterPath or FilterValues.
+// It is an error to pass an unfiltered Ignore option to Equal.
+func Ignore() Option { return ignore{} }
+
+type ignore struct{ core }
+
+func (ignore) isFiltered() bool                                                     { return false }
+func (ignore) filter(_ *state, _, _ reflect.Value, _ reflect.Type) applicableOption { return ignore{} }
+func (ignore) apply(_ *state, _, _ reflect.Value) bool                              { return true }
+func (ignore) String() string                                                       { return "Ignore()" }
+
+// invalid is a sentinel Option type to indicate that some options could not
+// be evaluated due to unexported fields.
+type invalid struct{ core }
+
+func (invalid) filter(_ *state, _, _ reflect.Value, _ reflect.Type) applicableOption { return invalid{} }
+func (invalid) apply(s *state, _, _ reflect.Value) bool {
+	const help = "consider using AllowUnexported or cmpopts.IgnoreUnexported"
+	panic(fmt.Sprintf("cannot handle unexported field: %#v\n%s", s.curPath, help))
+}
+
+// Transformer returns an Option that applies a transformation function that
+// converts values of a certain type into that of another.
+//
+// The transformer f must be a function "func(T) R" that converts values of
+// type T to those of type R and is implicitly filtered to input values
+// assignable to T. The transformer must not mutate T in any way.
+// If T and R are the same type, an additional filter must be applied to
+// act as the base case to prevent an infinite recursion applying the same
+// transform to itself (see the SortedSlice example).
+//
+// The name is a user provided label that is used as the Transform.Name in the
+// transformation PathStep. If empty, an arbitrary name is used.
+func Transformer(name string, f interface{}) Option {
+	v := reflect.ValueOf(f)
+	if !function.IsType(v.Type(), function.Transformer) || v.IsNil() {
+		panic(fmt.Sprintf("invalid transformer function: %T", f))
+	}
+	if name == "" {
+		name = "λ" // Lambda-symbol as place-holder for anonymous transformer
+	}
+	if !isValid(name) {
+		panic(fmt.Sprintf("invalid name: %q", name))
+	}
+	tr := &transformer{name: name, fnc: reflect.ValueOf(f)}
+	if ti := v.Type().In(0); ti.Kind() != reflect.Interface || ti.NumMethod() > 0 {
+		tr.typ = ti
+	}
+	return tr
+}
+
+type transformer struct {
+	core
+	name string
+	typ  reflect.Type  // T
+	fnc  reflect.Value // func(T) R
+}
+
+func (tr *transformer) isFiltered() bool { return tr.typ != nil }
+
+func (tr *transformer) filter(_ *state, _, _ reflect.Value, t reflect.Type) applicableOption {
+	if tr.typ == nil || t.AssignableTo(tr.typ) {
+		return tr
+	}
+	return nil
+}
+
+func (tr *transformer) apply(s *state, vx, vy reflect.Value) bool {
+	// Update path before calling the Transformer so that dynamic checks
+	// will use the updated path.
+	s.curPath.push(&transform{pathStep{tr.fnc.Type().Out(0)}, tr})
+	defer s.curPath.pop()
+
+	vx = s.callTRFunc(tr.fnc, vx)
+	vy = s.callTRFunc(tr.fnc, vy)
+	s.compareAny(vx, vy)
+	return true
+}
+
+func (tr transformer) String() string {
+	return fmt.Sprintf("Transformer(%s, %s)", tr.name, getFuncName(tr.fnc.Pointer()))
+}
+
+// Comparer returns an Option that determines whether two values are equal
+// to each other.
+//
+// The comparer f must be a function "func(T, T) bool" and is implicitly
+// filtered to input values assignable to T. If T is an interface, it is
+// possible that f is called with two values of different concrete types that
+// both implement T.
+//
+// The equality function must be:
+//	• Symmetric: equal(x, y) == equal(y, x)
+//	• Deterministic: equal(x, y) == equal(x, y)
+//	• Pure: equal(x, y) does not modify x or y
+func Comparer(f interface{}) Option {
+	v := reflect.ValueOf(f)
+	if !function.IsType(v.Type(), function.Equal) || v.IsNil() {
+		panic(fmt.Sprintf("invalid comparer function: %T", f))
+	}
+	cm := &comparer{fnc: v}
+	if ti := v.Type().In(0); ti.Kind() != reflect.Interface || ti.NumMethod() > 0 {
+		cm.typ = ti
+	}
+	return cm
+}
+
+type comparer struct {
+	core
+	typ reflect.Type  // T
+	fnc reflect.Value // func(T, T) bool
+}
+
+func (cm *comparer) isFiltered() bool { return cm.typ != nil }
+
+func (cm *comparer) filter(_ *state, _, _ reflect.Value, t reflect.Type) applicableOption {
+	if cm.typ == nil || t.AssignableTo(cm.typ) {
+		return cm
+	}
+	return nil
+}
+
+func (cm *comparer) apply(s *state, vx, vy reflect.Value) bool {
+	eq := s.callTTBFunc(cm.fnc, vx, vy)
+	s.report(eq, vx, vy)
+	return true
+}
+
+func (cm comparer) String() string {
+	return fmt.Sprintf("Comparer(%s)", getFuncName(cm.fnc.Pointer()))
+}
+
+// AllowUnexported returns an Option that forcibly allows operations on
+// unexported fields in certain structs, which are specified by passing in a
+// value of each struct type.
+//
+// Users of this option must understand that comparing on unexported fields
+// from external packages is not safe since changes in the internal
+// implementation of some external package may cause the result of Equal
+// to unexpectedly change. However, it may be valid to use this option on types
+// defined in an internal package where the semantic meaning of an unexported
+// field is in the control of the user.
+//
+// For some cases, a custom Comparer should be used instead that defines
+// equality as a function of the public API of a type rather than the underlying
+// unexported implementation.
+//
+// For example, the reflect.Type documentation defines equality to be determined
+// by the == operator on the interface (essentially performing a shallow pointer
+// comparison) and most attempts to compare *regexp.Regexp types are interested
+// in only checking that the regular expression strings are equal.
+// Both of these are accomplished using Comparers:
+//
+//	Comparer(func(x, y reflect.Type) bool { return x == y })
+//	Comparer(func(x, y *regexp.Regexp) bool { return x.String() == y.String() })
+//
+// In other cases, the cmpopts.IgnoreUnexported option can be used to ignore
+// all unexported fields on specified struct types.
+func AllowUnexported(types ...interface{}) Option {
+	if !supportAllowUnexported {
+		panic("AllowUnexported is not supported on App Engine Classic or GopherJS")
+	}
+	m := make(map[reflect.Type]bool)
+	for _, typ := range types {
+		t := reflect.TypeOf(typ)
+		if t.Kind() != reflect.Struct {
+			panic(fmt.Sprintf("invalid struct type: %T", typ))
+		}
+		m[t] = true
+	}
+	return visibleStructs(m)
+}
+
+type visibleStructs map[reflect.Type]bool
+
+func (visibleStructs) filter(_ *state, _, _ reflect.Value, _ reflect.Type) applicableOption {
+	panic("not implemented")
+}
+
+// reporter is an Option that configures how differences are reported.
+type reporter interface {
+	// TODO: Not exported yet.
+	//
+	// Perhaps add PushStep and PopStep and change Report to only accept
+	// a PathStep instead of the full-path? Adding a PushStep and PopStep makes
+	// it clear that we are traversing the value tree in a depth-first-search
+	// manner, which has an effect on how values are printed.
+
+	Option
+
+	// Report is called for every comparison made and will be provided with
+	// the two values being compared, the equality result, and the
+	// current path in the value tree. It is possible for x or y to be an
+	// invalid reflect.Value if one of the values is non-existent;
+	// which is possible with maps and slices.
+	Report(x, y reflect.Value, eq bool, p Path)
+}
+
+// normalizeOption normalizes the input options such that all Options groups
+// are flattened and groups with a single element are reduced to that element.
+// Only coreOptions and Options containing coreOptions are allowed.
+func normalizeOption(src Option) Option {
+	switch opts := flattenOptions(nil, Options{src}); len(opts) {
+	case 0:
+		return nil
+	case 1:
+		return opts[0]
+	default:
+		return opts
+	}
+}
+
+// flattenOptions copies all options in src to dst as a flat list.
+// Only coreOptions and Options containing coreOptions are allowed.
+func flattenOptions(dst, src Options) Options {
+	for _, opt := range src {
+		switch opt := opt.(type) {
+		case nil:
+			continue
+		case Options:
+			dst = flattenOptions(dst, opt)
+		case coreOption:
+			dst = append(dst, opt)
+		default:
+			panic(fmt.Sprintf("invalid option type: %T", opt))
+		}
+	}
+	return dst
+}
+
+// getFuncName returns a short function name from the pointer.
+// The string parsing logic works up until Go1.9.
+func getFuncName(p uintptr) string {
+	fnc := runtime.FuncForPC(p)
+	if fnc == nil {
+		return "<unknown>"
+	}
+	name := fnc.Name() // E.g., "long/path/name/mypkg.(mytype).(long/path/name/mypkg.myfunc)-fm"
+	if strings.HasSuffix(name, ")-fm") || strings.HasSuffix(name, ")·fm") {
+		// Strip the package name from method name.
+		name = strings.TrimSuffix(name, ")-fm")
+		name = strings.TrimSuffix(name, ")·fm")
+		if i := strings.LastIndexByte(name, '('); i >= 0 {
+			methodName := name[i+1:] // E.g., "long/path/name/mypkg.myfunc"
+			if j := strings.LastIndexByte(methodName, '.'); j >= 0 {
+				methodName = methodName[j+1:] // E.g., "myfunc"
+			}
+			name = name[:i] + methodName // E.g., "long/path/name/mypkg.(mytype)." + "myfunc"
+		}
+	}
+	if i := strings.LastIndexByte(name, '/'); i >= 0 {
+		// Strip the package name.
+		name = name[i+1:] // E.g., "mypkg.(mytype).myfunc"
+	}
+	return name
+}
diff --git a/vendor/github.com/google/go-cmp/cmp/path.go b/vendor/github.com/google/go-cmp/cmp/path.go
new file mode 100644
index 000000000..0c2eb333f
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/path.go
@@ -0,0 +1,293 @@
+// Copyright 2017, 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.md file.
+
+package cmp
+
+import (
+	"fmt"
+	"reflect"
+	"strings"
+	"unicode"
+	"unicode/utf8"
+)
+
+type (
+	// Path is a list of PathSteps describing the sequence of operations to get
+	// from some root type to the current position in the value tree.
+	// The first Path element is always an operation-less PathStep that exists
+	// simply to identify the initial type.
+	//
+	// When traversing structs with embedded structs, the embedded struct will
+	// always be accessed as a field before traversing the fields of the
+	// embedded struct themselves. That is, an exported field from the
+	// embedded struct will never be accessed directly from the parent struct.
+	Path []PathStep
+
+	// PathStep is a union-type for specific operations to traverse
+	// a value's tree structure. Users of this package never need to implement
+	// these types as values of this type will be returned by this package.
+	PathStep interface {
+		String() string
+		Type() reflect.Type // Resulting type after performing the path step
+		isPathStep()
+	}
+
+	// SliceIndex is an index operation on a slice or array at some index Key.
+	SliceIndex interface {
+		PathStep
+		Key() int // May return -1 if in a split state
+
+		// SplitKeys returns the indexes for indexing into slices in the
+		// x and y values, respectively. These indexes may differ due to the
+		// insertion or removal of an element in one of the slices, causing
+		// all of the indexes to be shifted. If an index is -1, then that
+		// indicates that the element does not exist in the associated slice.
+		//
+		// Key is guaranteed to return -1 if and only if the indexes returned
+		// by SplitKeys are not the same. SplitKeys will never return -1 for
+		// both indexes.
+		SplitKeys() (x int, y int)
+
+		isSliceIndex()
+	}
+	// MapIndex is an index operation on a map at some index Key.
+	MapIndex interface {
+		PathStep
+		Key() reflect.Value
+		isMapIndex()
+	}
+	// TypeAssertion represents a type assertion on an interface.
+	TypeAssertion interface {
+		PathStep
+		isTypeAssertion()
+	}
+	// StructField represents a struct field access on a field called Name.
+	StructField interface {
+		PathStep
+		Name() string
+		Index() int
+		isStructField()
+	}
+	// Indirect represents pointer indirection on the parent type.
+	Indirect interface {
+		PathStep
+		isIndirect()
+	}
+	// Transform is a transformation from the parent type to the current type.
+	Transform interface {
+		PathStep
+		Name() string
+		Func() reflect.Value
+		isTransform()
+	}
+)
+
+func (pa *Path) push(s PathStep) {
+	*pa = append(*pa, s)
+}
+
+func (pa *Path) pop() {
+	*pa = (*pa)[:len(*pa)-1]
+}
+
+// Last returns the last PathStep in the Path.
+// If the path is empty, this returns a non-nil PathStep that reports a nil Type.
+func (pa Path) Last() PathStep {
+	if len(pa) > 0 {
+		return pa[len(pa)-1]
+	}
+	return pathStep{}
+}
+
+// String returns the simplified path to a node.
+// The simplified path only contains struct field accesses.
+//
+// For example:
+//	MyMap.MySlices.MyField
+func (pa Path) String() string {
+	var ss []string
+	for _, s := range pa {
+		if _, ok := s.(*structField); ok {
+			ss = append(ss, s.String())
+		}
+	}
+	return strings.TrimPrefix(strings.Join(ss, ""), ".")
+}
+
+// GoString returns the path to a specific node using Go syntax.
+//
+// For example:
+//	(*root.MyMap["key"].(*mypkg.MyStruct).MySlices)[2][3].MyField
+func (pa Path) GoString() string {
+	var ssPre, ssPost []string
+	var numIndirect int
+	for i, s := range pa {
+		var nextStep PathStep
+		if i+1 < len(pa) {
+			nextStep = pa[i+1]
+		}
+		switch s := s.(type) {
+		case *indirect:
+			numIndirect++
+			pPre, pPost := "(", ")"
+			switch nextStep.(type) {
+			case *indirect:
+				continue // Next step is indirection, so let them batch up
+			case *structField:
+				numIndirect-- // Automatic indirection on struct fields
+			case nil:
+				pPre, pPost = "", "" // Last step; no need for parenthesis
+			}
+			if numIndirect > 0 {
+				ssPre = append(ssPre, pPre+strings.Repeat("*", numIndirect))
+				ssPost = append(ssPost, pPost)
+			}
+			numIndirect = 0
+			continue
+		case *transform:
+			ssPre = append(ssPre, s.trans.name+"(")
+			ssPost = append(ssPost, ")")
+			continue
+		case *typeAssertion:
+			// Elide type assertions immediately following a transform to
+			// prevent overly verbose path printouts.
+			// Some transforms return interface{} because of Go's lack of
+			// generics, but typically take in and return the exact same
+			// concrete type. Other times, the transform creates an anonymous
+			// struct, which will be very verbose to print.
+			if _, ok := nextStep.(*transform); ok {
+				continue
+			}
+		}
+		ssPost = append(ssPost, s.String())
+	}
+	for i, j := 0, len(ssPre)-1; i < j; i, j = i+1, j-1 {
+		ssPre[i], ssPre[j] = ssPre[j], ssPre[i]
+	}
+	return strings.Join(ssPre, "") + strings.Join(ssPost, "")
+}
+
+type (
+	pathStep struct {
+		typ reflect.Type
+	}
+
+	sliceIndex struct {
+		pathStep
+		xkey, ykey int
+	}
+	mapIndex struct {
+		pathStep
+		key reflect.Value
+	}
+	typeAssertion struct {
+		pathStep
+	}
+	structField struct {
+		pathStep
+		name string
+		idx  int
+
+		// These fields are used for forcibly accessing an unexported field.
+		// pvx, pvy, and field are only valid if unexported is true.
+		unexported bool
+		force      bool                // Forcibly allow visibility
+		pvx, pvy   reflect.Value       // Parent values
+		field      reflect.StructField // Field information
+	}
+	indirect struct {
+		pathStep
+	}
+	transform struct {
+		pathStep
+		trans *transformer
+	}
+)
+
+func (ps pathStep) Type() reflect.Type { return ps.typ }
+func (ps pathStep) String() string {
+	if ps.typ == nil {
+		return "<nil>"
+	}
+	s := ps.typ.String()
+	if s == "" || strings.ContainsAny(s, "{}\n") {
+		return "root" // Type too simple or complex to print
+	}
+	return fmt.Sprintf("{%s}", s)
+}
+
+func (si sliceIndex) String() string {
+	switch {
+	case si.xkey == si.ykey:
+		return fmt.Sprintf("[%d]", si.xkey)
+	case si.ykey == -1:
+		// [5->?] means "I don't know where X[5] went"
+		return fmt.Sprintf("[%d->?]", si.xkey)
+	case si.xkey == -1:
+		// [?->3] means "I don't know where Y[3] came from"
+		return fmt.Sprintf("[?->%d]", si.ykey)
+	default:
+		// [5->3] means "X[5] moved to Y[3]"
+		return fmt.Sprintf("[%d->%d]", si.xkey, si.ykey)
+	}
+}
+func (mi mapIndex) String() string      { return fmt.Sprintf("[%#v]", mi.key) }
+func (ta typeAssertion) String() string { return fmt.Sprintf(".(%v)", ta.typ) }
+func (sf structField) String() string   { return fmt.Sprintf(".%s", sf.name) }
+func (in indirect) String() string      { return "*" }
+func (tf transform) String() string     { return fmt.Sprintf("%s()", tf.trans.name) }
+
+func (si sliceIndex) Key() int {
+	if si.xkey != si.ykey {
+		return -1
+	}
+	return si.xkey
+}
+func (si sliceIndex) SplitKeys() (x, y int) { return si.xkey, si.ykey }
+func (mi mapIndex) Key() reflect.Value      { return mi.key }
+func (sf structField) Name() string         { return sf.name }
+func (sf structField) Index() int           { return sf.idx }
+func (tf transform) Name() string           { return tf.trans.name }
+func (tf transform) Func() reflect.Value    { return tf.trans.fnc }
+
+func (pathStep) isPathStep()           {}
+func (sliceIndex) isSliceIndex()       {}
+func (mapIndex) isMapIndex()           {}
+func (typeAssertion) isTypeAssertion() {}
+func (structField) isStructField()     {}
+func (indirect) isIndirect()           {}
+func (transform) isTransform()         {}
+
+var (
+	_ SliceIndex    = sliceIndex{}
+	_ MapIndex      = mapIndex{}
+	_ TypeAssertion = typeAssertion{}
+	_ StructField   = structField{}
+	_ Indirect      = indirect{}
+	_ Transform     = transform{}
+
+	_ PathStep = sliceIndex{}
+	_ PathStep = mapIndex{}
+	_ PathStep = typeAssertion{}
+	_ PathStep = structField{}
+	_ PathStep = indirect{}
+	_ PathStep = transform{}
+)
+
+// isExported reports whether the identifier is exported.
+func isExported(id string) bool {
+	r, _ := utf8.DecodeRuneInString(id)
+	return unicode.IsUpper(r)
+}
+
+// isValid reports whether the identifier is valid.
+// Empty and underscore-only strings are not valid.
+func isValid(id string) bool {
+	ok := id != "" && id != "_"
+	for j, c := range id {
+		ok = ok && (j > 0 || !unicode.IsDigit(c))
+		ok = ok && (c == '_' || unicode.IsLetter(c) || unicode.IsDigit(c))
+	}
+	return ok
+}
diff --git a/vendor/github.com/google/go-cmp/cmp/reporter.go b/vendor/github.com/google/go-cmp/cmp/reporter.go
new file mode 100644
index 000000000..a21d0cded
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/reporter.go
@@ -0,0 +1,53 @@
+// Copyright 2017, 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.md file.
+
+package cmp
+
+import (
+	"fmt"
+	"reflect"
+	"strings"
+
+	"github.com/google/go-cmp/cmp/internal/value"
+)
+
+type defaultReporter struct {
+	Option
+	diffs  []string // List of differences, possibly truncated
+	ndiffs int      // Total number of differences
+	nbytes int      // Number of bytes in diffs
+	nlines int      // Number of lines in diffs
+}
+
+var _ reporter = (*defaultReporter)(nil)
+
+func (r *defaultReporter) Report(x, y reflect.Value, eq bool, p Path) {
+	if eq {
+		return // Ignore equal results
+	}
+	const maxBytes = 4096
+	const maxLines = 256
+	r.ndiffs++
+	if r.nbytes < maxBytes && r.nlines < maxLines {
+		sx := value.Format(x, true)
+		sy := value.Format(y, true)
+		if sx == sy {
+			// Stringer is not helpful, so rely on more exact formatting.
+			sx = value.Format(x, false)
+			sy = value.Format(y, false)
+		}
+		s := fmt.Sprintf("%#v:\n\t-: %s\n\t+: %s\n", p, sx, sy)
+		r.diffs = append(r.diffs, s)
+		r.nbytes += len(s)
+		r.nlines += strings.Count(s, "\n")
+	}
+}
+
+func (r *defaultReporter) String() string {
+	s := strings.Join(r.diffs, "")
+	if r.ndiffs == len(r.diffs) {
+		return s
+	}
+	return fmt.Sprintf("%s... %d more differences ...", s, len(r.diffs)-r.ndiffs)
+}
diff --git a/vendor/github.com/google/go-cmp/cmp/unsafe_reflect.go b/vendor/github.com/google/go-cmp/cmp/unsafe_reflect.go
new file mode 100644
index 000000000..81fb82632
--- /dev/null
+++ b/vendor/github.com/google/go-cmp/cmp/unsafe_reflect.go
@@ -0,0 +1,23 @@
+// Copyright 2017, 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.md file.
+
+// +build !appengine,!js
+
+package cmp
+
+import (
+	"reflect"
+	"unsafe"
+)
+
+const supportAllowUnexported = true
+
+// unsafeRetrieveField uses unsafe to forcibly retrieve any field from a struct
+// such that the value has read-write permissions.
+//
+// The parent struct, v, must be addressable, while f must be a StructField
+// describing the field to retrieve.
+func unsafeRetrieveField(v reflect.Value, f reflect.StructField) reflect.Value {
+	return reflect.NewAt(f.Type, unsafe.Pointer(v.UnsafeAddr()+f.Offset)).Elem()
+}
diff --git a/vendor/vendor.json b/vendor/vendor.json
index 8012ed6cf..ab20a7786 100644
--- a/vendor/vendor.json
+++ b/vendor/vendor.json
@@ -408,14 +408,14 @@
 		{
 			"checksumSHA1": "rArZ5mYIe9I1L5PRQOJu8BwafFw=",
 			"path": "github.com/docker/docker/pkg/pools",
-			"revision": "52debcd58ac91bf68503ce60561536911b74ff05",
-			"revisionTime": "2016-05-20T15:17:10Z"
+			"revision": "da39e9a4f920a15683dd0f23923c302d4db6eed5",
+			"revisionTime": "2016-05-28T08:11:04Z"
 		},
 		{
 			"checksumSHA1": "rArZ5mYIe9I1L5PRQOJu8BwafFw=",
 			"path": "github.com/docker/docker/pkg/pools",
-			"revision": "da39e9a4f920a15683dd0f23923c302d4db6eed5",
-			"revisionTime": "2016-05-28T08:11:04Z"
+			"revision": "52debcd58ac91bf68503ce60561536911b74ff05",
+			"revisionTime": "2016-05-20T15:17:10Z"
 		},
 		{
 			"checksumSHA1": "txf3EORYff4hO6PEvwBm2lyh1MU=",
@@ -447,18 +447,18 @@
 			"revision": "518945b6bd30dc535addf2aac9d5eeefbfc5e523",
 			"revisionTime": "2017-01-10T19:35:11Z"
 		},
-		{
-			"checksumSHA1": "jtruj9m4YJtPaXqUZ5PXdgFQjRA=",
-			"path": "github.com/docker/docker/pkg/system",
-			"revision": "47cd4bf526723100bdbbd5a6bb983cdd668882e9",
-			"revisionTime": "2017-03-13T17:42:22Z"
-		},
 		{
 			"checksumSHA1": "NGqzajRG0Vri8ConQuHY3cCs1RA=",
 			"path": "github.com/docker/docker/pkg/system",
 			"revision": "fe9ab0588606a5566d065bc68ae68f3926ddaa72",
 			"revisionTime": "2017-02-28T18:28:19Z"
 		},
+		{
+			"checksumSHA1": "jtruj9m4YJtPaXqUZ5PXdgFQjRA=",
+			"path": "github.com/docker/docker/pkg/system",
+			"revision": "47cd4bf526723100bdbbd5a6bb983cdd668882e9",
+			"revisionTime": "2017-03-13T17:42:22Z"
+		},
 		{
 			"checksumSHA1": "Nfp/0W+HK8ZEgklbSWmjJGXTJew=",
 			"path": "github.com/docker/docker/pkg/term",
@@ -598,6 +598,36 @@
 			"revision": "d9eb7a3d35ec988b8585d4a0068e462c27d28380",
 			"revisionTime": "2016-05-29T05:00:41Z"
 		},
+		{
+			"checksumSHA1": "+suAHHPBmbdZf/HusugaL4/H+NE=",
+			"path": "github.com/google/go-cmp/cmp",
+			"revision": "d5735f74713c51f7450a43d0a98d41ce2c1db3cb",
+			"revisionTime": "2017-09-01T21:42:48Z"
+		},
+		{
+			"checksumSHA1": "VmBLfV9TChrjNu8Z96wZkYie1aI=",
+			"path": "github.com/google/go-cmp/cmp/cmpopts",
+			"revision": "d5735f74713c51f7450a43d0a98d41ce2c1db3cb",
+			"revisionTime": "2017-09-01T21:42:48Z"
+		},
+		{
+			"checksumSHA1": "eTwchtMX+RMJUvle2wt295P2h10=",
+			"path": "github.com/google/go-cmp/cmp/internal/diff",
+			"revision": "d5735f74713c51f7450a43d0a98d41ce2c1db3cb",
+			"revisionTime": "2017-09-01T21:42:48Z"
+		},
+		{
+			"checksumSHA1": "kYtvRhMjM0X4bvEjR3pqEHLw1qo=",
+			"path": "github.com/google/go-cmp/cmp/internal/function",
+			"revision": "d5735f74713c51f7450a43d0a98d41ce2c1db3cb",
+			"revisionTime": "2017-09-01T21:42:48Z"
+		},
+		{
+			"checksumSHA1": "f+mgZLvc4VITtMmBv0bmew4rL2Y=",
+			"path": "github.com/google/go-cmp/cmp/internal/value",
+			"revision": "d5735f74713c51f7450a43d0a98d41ce2c1db3cb",
+			"revisionTime": "2017-09-01T21:42:48Z"
+		},
 		{
 			"comment": "1.0.0",
 			"path": "github.com/gorhill/cronexpr",