Update dependecies (#19635)

vendor/github.com/aws/aws-sdk-go/aws/endpoints/defaults.go

@@ -1669,6 +1669,147 @@ var awsPartition = partition{
 				},
 			},
 		},
+		"api.tunneling.iot": service{
+			Defaults: endpointDefaults{
+				defaultKey{}: endpoint{},
+				defaultKey{
+					Variant: fipsVariant,
+				}: endpoint{
+					Hostname: "api.tunneling.iot-fips.{region}.{dnsSuffix}",
+				},
+			},
+			Endpoints: serviceEndpoints{
+				endpointKey{
+					Region: "ap-east-1",
+				}: endpoint{},
+				endpointKey{
+					Region: "ap-northeast-1",
+				}: endpoint{},
+				endpointKey{
+					Region: "ap-northeast-2",
+				}: endpoint{},
+				endpointKey{
+					Region: "ap-south-1",
+				}: endpoint{},
+				endpointKey{
+					Region: "ap-southeast-1",
+				}: endpoint{},
+				endpointKey{
+					Region: "ap-southeast-2",
+				}: endpoint{},
+				endpointKey{
+					Region: "ca-central-1",
+				}: endpoint{},
+				endpointKey{
+					Region:  "ca-central-1",
+					Variant: fipsVariant,
+				}: endpoint{
+					Hostname: "api.tunneling.iot-fips.ca-central-1.amazonaws.com",
+				},
+				endpointKey{
+					Region: "eu-central-1",
+				}: endpoint{},
+				endpointKey{
+					Region: "eu-north-1",
+				}: endpoint{},
+				endpointKey{
+					Region: "eu-west-1",
+				}: endpoint{},
+				endpointKey{
+					Region: "eu-west-2",
+				}: endpoint{},
+				endpointKey{
+					Region: "eu-west-3",
+				}: endpoint{},
+				endpointKey{
+					Region: "fips-ca-central-1",
+				}: endpoint{
+					Hostname: "api.tunneling.iot-fips.ca-central-1.amazonaws.com",
+					CredentialScope: credentialScope{
+						Region: "ca-central-1",
+					},
+					Deprecated: boxedTrue,
+				},
+				endpointKey{
+					Region: "fips-us-east-1",
+				}: endpoint{
+					Hostname: "api.tunneling.iot-fips.us-east-1.amazonaws.com",
+					CredentialScope: credentialScope{
+						Region: "us-east-1",
+					},
+					Deprecated: boxedTrue,
+				},
+				endpointKey{
+					Region: "fips-us-east-2",
+				}: endpoint{
+					Hostname: "api.tunneling.iot-fips.us-east-2.amazonaws.com",
+					CredentialScope: credentialScope{
+						Region: "us-east-2",
+					},
+					Deprecated: boxedTrue,
+				},
+				endpointKey{
+					Region: "fips-us-west-1",
+				}: endpoint{
+					Hostname: "api.tunneling.iot-fips.us-west-1.amazonaws.com",
+					CredentialScope: credentialScope{
+						Region: "us-west-1",
+					},
+					Deprecated: boxedTrue,
+				},
+				endpointKey{
+					Region: "fips-us-west-2",
+				}: endpoint{
+					Hostname: "api.tunneling.iot-fips.us-west-2.amazonaws.com",
+					CredentialScope: credentialScope{
+						Region: "us-west-2",
+					},
+					Deprecated: boxedTrue,
+				},
+				endpointKey{
+					Region: "me-south-1",
+				}: endpoint{},
+				endpointKey{
+					Region: "sa-east-1",
+				}: endpoint{},
+				endpointKey{
+					Region: "us-east-1",
+				}: endpoint{},
+				endpointKey{
+					Region:  "us-east-1",
+					Variant: fipsVariant,
+				}: endpoint{
+					Hostname: "api.tunneling.iot-fips.us-east-1.amazonaws.com",
+				},
+				endpointKey{
+					Region: "us-east-2",
+				}: endpoint{},
+				endpointKey{
+					Region:  "us-east-2",
+					Variant: fipsVariant,
+				}: endpoint{
+					Hostname: "api.tunneling.iot-fips.us-east-2.amazonaws.com",
+				},
+				endpointKey{
+					Region: "us-west-1",
+				}: endpoint{},
+				endpointKey{
+					Region:  "us-west-1",
+					Variant: fipsVariant,
+				}: endpoint{
+					Hostname: "api.tunneling.iot-fips.us-west-1.amazonaws.com",
+				},
+				endpointKey{
+					Region: "us-west-2",
+				}: endpoint{},
+				endpointKey{
+					Region:  "us-west-2",
+					Variant: fipsVariant,
+				}: endpoint{
+					Hostname: "api.tunneling.iot-fips.us-west-2.amazonaws.com",
+				},
+			},
+		},
 		"apigateway": service{
 			Endpoints: serviceEndpoints{
 				endpointKey{
@@ -9817,6 +9958,9 @@ var awsPartition = partition{
 				endpointKey{
 					Region: "ap-southeast-2",
 				}: endpoint{},
+				endpointKey{
+					Region: "ap-southeast-3",
+				}: endpoint{},
 				endpointKey{
 					Region: "ca-central-1",
 				}: endpoint{},
@@ -13547,6 +13691,9 @@ var awsPartition = partition{
 		},
 		"mq": service{
 			Endpoints: serviceEndpoints{
+				endpointKey{
+					Region: "af-south-1",
+				}: endpoint{},
 				endpointKey{
 					Region: "ap-east-1",
 				}: endpoint{},
@@ -20669,6 +20816,9 @@ var awsPartition = partition{
 				endpointKey{
 					Region: "ap-northeast-2",
 				}: endpoint{},
+				endpointKey{
+					Region: "ap-northeast-3",
+				}: endpoint{},
 				endpointKey{
 					Region: "ap-south-1",
 				}: endpoint{},
@@ -21954,6 +22104,16 @@ var awscnPartition = partition{
 				}: endpoint{},
 			},
 		},
+		"api.tunneling.iot": service{
+			Endpoints: serviceEndpoints{
+				endpointKey{
+					Region: "cn-north-1",
+				}: endpoint{},
+				endpointKey{
+					Region: "cn-northwest-1",
+				}: endpoint{},
+			},
+		},
 		"apigateway": service{
 			Endpoints: serviceEndpoints{
 				endpointKey{
@@ -23462,6 +23622,14 @@ var awsusgovPartition = partition{
 			},
 		},
 		"acm": service{
+			Defaults: endpointDefaults{
+				defaultKey{}: endpoint{},
+				defaultKey{
+					Variant: fipsVariant,
+				}: endpoint{
+					Hostname: "acm.{region}.{dnsSuffix}",
+				},
+			},
 			Endpoints: serviceEndpoints{
 				endpointKey{
 					Region: "us-gov-east-1",
@@ -23752,6 +23920,54 @@ var awsusgovPartition = partition{
 				},
 			},
 		},
+		"api.tunneling.iot": service{
+			Defaults: endpointDefaults{
+				defaultKey{}: endpoint{},
+				defaultKey{
+					Variant: fipsVariant,
+				}: endpoint{
+					Hostname: "api.tunneling.iot-fips.{region}.{dnsSuffix}",
+				},
+			},
+			Endpoints: serviceEndpoints{
+				endpointKey{
+					Region: "fips-us-gov-east-1",
+				}: endpoint{
+					Hostname: "api.tunneling.iot-fips.us-gov-east-1.amazonaws.com",
+					CredentialScope: credentialScope{
+						Region: "us-gov-east-1",
+					},
+					Deprecated: boxedTrue,
+				},
+				endpointKey{
+					Region: "fips-us-gov-west-1",
+				}: endpoint{
+					Hostname: "api.tunneling.iot-fips.us-gov-west-1.amazonaws.com",
+					CredentialScope: credentialScope{
+						Region: "us-gov-west-1",
+					},
+					Deprecated: boxedTrue,
+				},
+				endpointKey{
+					Region: "us-gov-east-1",
+				}: endpoint{},
+				endpointKey{
+					Region:  "us-gov-east-1",
+					Variant: fipsVariant,
+				}: endpoint{
+					Hostname: "api.tunneling.iot-fips.us-gov-east-1.amazonaws.com",
+				},
+				endpointKey{
+					Region: "us-gov-west-1",
+				}: endpoint{},
+				endpointKey{
+					Region:  "us-gov-west-1",
+					Variant: fipsVariant,
+				}: endpoint{
+					Hostname: "api.tunneling.iot-fips.us-gov-west-1.amazonaws.com",
+				},
+			},
+		},
 		"apigateway": service{
 			Endpoints: serviceEndpoints{
 				endpointKey{
@@ -24688,6 +24904,14 @@ var awsusgovPartition = partition{
 			},
 		},
 		"ec2": service{
+			Defaults: endpointDefaults{
+				defaultKey{}: endpoint{},
+				defaultKey{
+					Variant: fipsVariant,
+				}: endpoint{
+					Hostname: "ec2.{region}.{dnsSuffix}",
+				},
+			},
 			Endpoints: serviceEndpoints{
 				endpointKey{
 					Region: "us-gov-east-1",

vendor/github.com/aws/aws-sdk-go/aws/version.go

@@ -5,4 +5,4 @@ package aws
 const SDKName = "aws-sdk-go"
 
 // SDKVersion is the version of this SDK
-const SDKVersion = "1.42.49"
+const SDKVersion = "1.43.6"

vendor/github.com/aws/aws-sdk-go/private/protocol/rest/build.go

@@ -272,6 +272,9 @@ func convertType(v reflect.Value, tag reflect.StructTag) (str string, err error)
 
 	switch value := v.Interface().(type) {
 	case string:
+		if tag.Get("suppressedJSONValue") == "true" && tag.Get("location") == "header" {
+			value = base64.StdEncoding.EncodeToString([]byte(value))
+		}
 		str = value
 	case []byte:
 		str = base64.StdEncoding.EncodeToString(value)
@@ -306,5 +309,6 @@ func convertType(v reflect.Value, tag reflect.StructTag) (str string, err error)
 		err := fmt.Errorf("unsupported value for param %v (%s)", v.Interface(), v.Type())
 		return "", err
 	}
+
 	return str, nil
 }

vendor/github.com/aws/aws-sdk-go/private/protocol/rest/unmarshal.go

@@ -204,6 +204,13 @@ func unmarshalHeader(v reflect.Value, header string, tag reflect.StructTag) erro
 
 	switch v.Interface().(type) {
 	case *string:
+		if tag.Get("suppressedJSONValue") == "true" && tag.Get("location") == "header" {
+			b, err := base64.StdEncoding.DecodeString(header)
+			if err != nil {
+				return fmt.Errorf("failed to decode JSONValue, %v", err)
+			}
+			header = string(b)
+		}
 		v.Set(reflect.ValueOf(&header))
 	case []byte:
 		b, err := base64.StdEncoding.DecodeString(header)

vendor/github.com/blevesearch/bleve/v2/go.mod

@@ -12,11 +12,11 @@ require (
 	github.com/blevesearch/snowballstem v0.9.0
 	github.com/blevesearch/upsidedown_store_api v1.0.1
 	github.com/blevesearch/vellum v1.0.7
-	github.com/blevesearch/zapx/v11 v11.3.2
-	github.com/blevesearch/zapx/v12 v12.3.2
-	github.com/blevesearch/zapx/v13 v13.3.2
-	github.com/blevesearch/zapx/v14 v14.3.2
-	github.com/blevesearch/zapx/v15 v15.3.2
+	github.com/blevesearch/zapx/v11 v11.3.3
+	github.com/blevesearch/zapx/v12 v12.3.3
+	github.com/blevesearch/zapx/v13 v13.3.3
+	github.com/blevesearch/zapx/v14 v14.3.3
+	github.com/blevesearch/zapx/v15 v15.3.3
 	github.com/couchbase/moss v0.2.0
 	github.com/golang/protobuf v1.3.2
 	github.com/kljensen/snowball v0.6.0

vendor/github.com/blevesearch/bleve/v2/go.sum

@@ -21,16 +21,16 @@ github.com/blevesearch/upsidedown_store_api v1.0.1 h1:1SYRwyoFLwG3sj0ed89RLtM15a
 github.com/blevesearch/upsidedown_store_api v1.0.1/go.mod h1:MQDVGpHZrpe3Uy26zJBf/a8h0FZY6xJbthIMm8myH2Q=
 github.com/blevesearch/vellum v1.0.7 h1:+vn8rfyCRHxKVRgDLeR0FAXej2+6mEb5Q15aQE/XESQ=
 github.com/blevesearch/vellum v1.0.7/go.mod h1:doBZpmRhwTsASB4QdUZANlJvqVAUdUyX0ZK7QJCTeBE=
-github.com/blevesearch/zapx/v11 v11.3.2 h1:TDdcbaA0Yz3Y5zpTrpvyW1AeicqWTJL3g8D5g48RiHM=
-github.com/blevesearch/zapx/v11 v11.3.2/go.mod h1:YzTfUm4kS3e8OmTXDHVV8OzC5MWPO/VPJZQgPNVb4Lc=
-github.com/blevesearch/zapx/v12 v12.3.2 h1:XB09XMg/3ibeIJRCm2zjkaVwrtAuk6c55YRSmVlwUDk=
-github.com/blevesearch/zapx/v12 v12.3.2/go.mod h1:RMl6lOZqF+sTxKvhQDJ5yK2LT3Mu7E2p/jGdjAaiRxs=
-github.com/blevesearch/zapx/v13 v13.3.2 h1:mTvALh6oayreac07VRAv94FLvTHeSBM9sZ1gmVt0N2k=
-github.com/blevesearch/zapx/v13 v13.3.2/go.mod h1:eppobNM35U4C22yDvTuxV9xPqo10pwfP/jugL4INWG4=
-github.com/blevesearch/zapx/v14 v14.3.2 h1:oW36JVaZDzrzmBa1X5jdTIYzdhkOQnr/ie13Cb2X7MQ=
-github.com/blevesearch/zapx/v14 v14.3.2/go.mod h1:zXNcVzukh0AvG57oUtT1T0ndi09H0kELNaNmekEy0jw=
-github.com/blevesearch/zapx/v15 v15.3.2 h1:OZNE4CQ9hQhnB21ySC7x2/9Q35U3WtRXLAh5L2gdCXc=
-github.com/blevesearch/zapx/v15 v15.3.2/go.mod h1:C+f/97ZzTzK6vt/7sVlZdzZxKu+5+j4SrGCvr9dJzaY=
+github.com/blevesearch/zapx/v11 v11.3.3 h1:8vQMO5hdA2qPCmicIMuKS+qcvUAEh6Vcb0uve4Nh8e4=
+github.com/blevesearch/zapx/v11 v11.3.3/go.mod h1:YzTfUm4kS3e8OmTXDHVV8OzC5MWPO/VPJZQgPNVb4Lc=
+github.com/blevesearch/zapx/v12 v12.3.3 h1:MQO5YNI8MqdPz12ALCoXiJw5cl9QQamYZSp285Z/+Mo=
+github.com/blevesearch/zapx/v12 v12.3.3/go.mod h1:RMl6lOZqF+sTxKvhQDJ5yK2LT3Mu7E2p/jGdjAaiRxs=
+github.com/blevesearch/zapx/v13 v13.3.3 h1:TS4xpMK1ARPYHq+1WwuEOKMOiwvKpTK3RuWOkKlI7BE=
+github.com/blevesearch/zapx/v13 v13.3.3/go.mod h1:eppobNM35U4C22yDvTuxV9xPqo10pwfP/jugL4INWG4=
+github.com/blevesearch/zapx/v14 v14.3.3 h1:dqqAzGphKl0yehHKKntDHKlEMhi9B/tJrD4OsWpY7YE=
+github.com/blevesearch/zapx/v14 v14.3.3/go.mod h1:zXNcVzukh0AvG57oUtT1T0ndi09H0kELNaNmekEy0jw=
+github.com/blevesearch/zapx/v15 v15.3.3 h1:60oE+qsJkveLenJmbc0eaH59GWYCbJJsPDV6Z5hEoYY=
+github.com/blevesearch/zapx/v15 v15.3.3/go.mod h1:C+f/97ZzTzK6vt/7sVlZdzZxKu+5+j4SrGCvr9dJzaY=
 github.com/coreos/etcd v3.3.10+incompatible/go.mod h1:uF7uidLiAD3TWHmW31ZFd/JWoc32PjwdhPthX9715RE=
 github.com/coreos/go-etcd v2.0.0+incompatible/go.mod h1:Jez6KQU2B/sWsbdaef3ED8NzMklzPG4d5KIOhIy30Tk=
 github.com/coreos/go-semver v0.2.0/go.mod h1:nnelYz7RCh+5ahJtPPxZlU+153eP4D4r3EedlOD2RNk=

vendor/github.com/blevesearch/bleve/v2/index/scorch/introducer.go

@@ -16,6 +16,7 @@ package scorch
 
 import (
 	"fmt"
+	"path/filepath"
 	"sync/atomic"
 
 	"github.com/RoaringBitmap/roaring"
@@ -127,6 +128,7 @@ func (s *Scorch) introduceSegment(next *segmentIntroduction) error {
 	// iterate through current segments
 	var running uint64
 	var docsToPersistCount, memSegments, fileSegments uint64
+	var droppedSegmentFiles []string
 	for i := range root.segment {
 		// see if optimistic work included this segment
 		delta, ok := next.obsoletes[root.segment[i].id]
@@ -164,6 +166,9 @@ func (s *Scorch) introduceSegment(next *segmentIntroduction) error {
 			root.segment[i].segment.AddRef()
 			newSnapshot.offsets = append(newSnapshot.offsets, running)
 			running += newss.segment.Count()
+		} else if seg, ok := newss.segment.(segment.PersistedSegment); ok {
+			droppedSegmentFiles = append(droppedSegmentFiles,
+				filepath.Base(seg.Path()))
 		}
 
 		if isMemorySegment(root.segment[i]) {
@@ -228,6 +233,12 @@ func (s *Scorch) introduceSegment(next *segmentIntroduction) error {
 		_ = rootPrev.DecRef()
 	}
 
+	// update the removal eligibility for those segment files
+	// that are not a part of the latest root.
+	for _, filename := range droppedSegmentFiles {
+		s.unmarkIneligibleForRemoval(filename)
+	}
+
 	close(next.applied)
 
 	return nil
@@ -332,6 +343,7 @@ func (s *Scorch) introduceMerge(nextMerge *segmentMerge) {
 	// iterate through current segments
 	newSegmentDeleted := roaring.NewBitmap()
 	var running, docsToPersistCount, memSegments, fileSegments uint64
+	var droppedSegmentFiles []string
 	for i := range root.segment {
 		segmentID := root.segment[i].id
 		if segSnapAtMerge, ok := nextMerge.old[segmentID]; ok {
@@ -374,8 +386,12 @@ func (s *Scorch) introduceMerge(nextMerge *segmentMerge) {
 			} else {
 				fileSegments++
 			}
+		} else if root.segment[i].LiveSize() == 0 {
+			if seg, ok := root.segment[i].segment.(segment.PersistedSegment); ok {
+				droppedSegmentFiles = append(droppedSegmentFiles,
+					filepath.Base(seg.Path()))
+			}
 		}
-
 	}
 
 	// before the newMerge introduction, need to clean the newly
@@ -441,6 +457,12 @@ func (s *Scorch) introduceMerge(nextMerge *segmentMerge) {
 		_ = rootPrev.DecRef()
 	}
 
+	// update the removal eligibility for those segment files
+	// that are not a part of the latest root.
+	for _, filename := range droppedSegmentFiles {
+		s.unmarkIneligibleForRemoval(filename)
+	}
+
 	// notify requester that we incorporated this
 	nextMerge.notifyCh <- &mergeTaskIntroStatus{
 		indexSnapshot: newSnapshot,

vendor/github.com/blevesearch/bleve/v2/search/facets_builder.go

@@ -127,6 +127,9 @@ type TermFacets struct {
 }
 
 func (tf *TermFacets) Terms() []*TermFacet {
+	if tf == nil {
+		return []*TermFacet{}
+	}
 	return tf.termFacets
 }
 

vendor/github.com/blevesearch/zapx/v11/memuvarint.go

@@ -39,6 +39,11 @@ func (r *memUvarintReader) Len() int {
 // ReadUvarint reads an encoded uint64.  The original code this was
 // based on is at encoding/binary/ReadUvarint().
 func (r *memUvarintReader) ReadUvarint() (uint64, error) {
+	if r.C >= len(r.S) {
+		// nothing else to read
+		return 0, nil
+	}
+
 	var x uint64
 	var s uint
 	var C = r.C
@@ -59,7 +64,7 @@ func (r *memUvarintReader) ReadUvarint() (uint64, error) {
 			// why the "extra" >= check?  The normal case is that s <
 			// 63, so we check this single >= guard first so that we
 			// hit the normal, nil-error return pathway sooner.
-			if s >= 63 && (s > 63 || s == 63 && b > 1) {
+			if s >= 63 && (s > 63 || b > 1) {
 				return 0, fmt.Errorf("memUvarintReader overflow")
 			}
 
@@ -74,6 +79,10 @@ func (r *memUvarintReader) ReadUvarint() (uint64, error) {
 // SkipUvarint skips ahead one encoded uint64.
 func (r *memUvarintReader) SkipUvarint() {
 	for {
+		if r.C >= len(r.S) {
+			return
+		}
+
 		b := r.S[r.C]
 		r.C++
 

vendor/github.com/blevesearch/zapx/v12/memuvarint.go

@@ -39,6 +39,11 @@ func (r *memUvarintReader) Len() int {
 // ReadUvarint reads an encoded uint64.  The original code this was
 // based on is at encoding/binary/ReadUvarint().
 func (r *memUvarintReader) ReadUvarint() (uint64, error) {
+	if r.C >= len(r.S) {
+		// nothing else to read
+		return 0, nil
+	}
+
 	var x uint64
 	var s uint
 	var C = r.C
@@ -59,7 +64,7 @@ func (r *memUvarintReader) ReadUvarint() (uint64, error) {
 			// why the "extra" >= check?  The normal case is that s <
 			// 63, so we check this single >= guard first so that we
 			// hit the normal, nil-error return pathway sooner.
-			if s >= 63 && (s > 63 || s == 63 && b > 1) {
+			if s >= 63 && (s > 63 || b > 1) {
 				return 0, fmt.Errorf("memUvarintReader overflow")
 			}
 
@@ -74,6 +79,10 @@ func (r *memUvarintReader) ReadUvarint() (uint64, error) {
 // SkipUvarint skips ahead one encoded uint64.
 func (r *memUvarintReader) SkipUvarint() {
 	for {
+		if r.C >= len(r.S) {
+			return
+		}
+
 		b := r.S[r.C]
 		r.C++
 

vendor/github.com/blevesearch/zapx/v13/memuvarint.go

@@ -39,6 +39,11 @@ func (r *memUvarintReader) Len() int {
 // ReadUvarint reads an encoded uint64.  The original code this was
 // based on is at encoding/binary/ReadUvarint().
 func (r *memUvarintReader) ReadUvarint() (uint64, error) {
+	if r.C >= len(r.S) {
+		// nothing else to read
+		return 0, nil
+	}
+
 	var x uint64
 	var s uint
 	var C = r.C
@@ -59,7 +64,7 @@ func (r *memUvarintReader) ReadUvarint() (uint64, error) {
 			// why the "extra" >= check?  The normal case is that s <
 			// 63, so we check this single >= guard first so that we
 			// hit the normal, nil-error return pathway sooner.
-			if s >= 63 && (s > 63 || s == 63 && b > 1) {
+			if s >= 63 && (s > 63 || b > 1) {
 				return 0, fmt.Errorf("memUvarintReader overflow")
 			}
 
@@ -74,6 +79,10 @@ func (r *memUvarintReader) ReadUvarint() (uint64, error) {
 // SkipUvarint skips ahead one encoded uint64.
 func (r *memUvarintReader) SkipUvarint() {
 	for {
+		if r.C >= len(r.S) {
+			return
+		}
+
 		b := r.S[r.C]
 		r.C++
 

vendor/github.com/blevesearch/zapx/v14/memuvarint.go

@@ -39,6 +39,11 @@ func (r *memUvarintReader) Len() int {
 // ReadUvarint reads an encoded uint64.  The original code this was
 // based on is at encoding/binary/ReadUvarint().
 func (r *memUvarintReader) ReadUvarint() (uint64, error) {
+	if r.C >= len(r.S) {
+		// nothing else to read
+		return 0, nil
+	}
+
 	var x uint64
 	var s uint
 	var C = r.C
@@ -59,7 +64,7 @@ func (r *memUvarintReader) ReadUvarint() (uint64, error) {
 			// why the "extra" >= check?  The normal case is that s <
 			// 63, so we check this single >= guard first so that we
 			// hit the normal, nil-error return pathway sooner.
-			if s >= 63 && (s > 63 || s == 63 && b > 1) {
+			if s >= 63 && (s > 63 || b > 1) {
 				return 0, fmt.Errorf("memUvarintReader overflow")
 			}
 
@@ -74,6 +79,10 @@ func (r *memUvarintReader) ReadUvarint() (uint64, error) {
 // SkipUvarint skips ahead one encoded uint64.
 func (r *memUvarintReader) SkipUvarint() {
 	for {
+		if r.C >= len(r.S) {
+			return
+		}
+
 		b := r.S[r.C]
 		r.C++
 

vendor/github.com/blevesearch/zapx/v15/memuvarint.go

@@ -39,6 +39,11 @@ func (r *memUvarintReader) Len() int {
 // ReadUvarint reads an encoded uint64.  The original code this was
 // based on is at encoding/binary/ReadUvarint().
 func (r *memUvarintReader) ReadUvarint() (uint64, error) {
+	if r.C >= len(r.S) {
+		// nothing else to read
+		return 0, nil
+	}
+
 	var x uint64
 	var s uint
 	var C = r.C
@@ -59,7 +64,7 @@ func (r *memUvarintReader) ReadUvarint() (uint64, error) {
 			// why the "extra" >= check?  The normal case is that s <
 			// 63, so we check this single >= guard first so that we
 			// hit the normal, nil-error return pathway sooner.
-			if s >= 63 && (s > 63 || s == 63 && b > 1) {
+			if s >= 63 && (s > 63 || b > 1) {
 				return 0, fmt.Errorf("memUvarintReader overflow")
 			}
 
@@ -74,6 +79,10 @@ func (r *memUvarintReader) ReadUvarint() (uint64, error) {
 // SkipUvarint skips ahead one encoded uint64.
 func (r *memUvarintReader) SkipUvarint() {
 	for {
+		if r.C >= len(r.S) {
+			return
+		}
+
 		b := r.S[r.C]
 		r.C++
 

vendor/github.com/gorilla/websocket/README.md

@@ -6,6 +6,13 @@
 Gorilla WebSocket is a [Go](http://golang.org/) implementation of the
 [WebSocket](http://www.rfc-editor.org/rfc/rfc6455.txt) protocol.
 
+
+---
+
+⚠️ **[The Gorilla WebSocket Package is looking for a new maintainer](https://github.com/gorilla/websocket/issues/370)**
+
+---
+
 ### Documentation
 
 * [API Reference](https://pkg.go.dev/github.com/gorilla/websocket?tab=doc)
@@ -30,35 +37,3 @@ The Gorilla WebSocket package passes the server tests in the [Autobahn Test
 Suite](https://github.com/crossbario/autobahn-testsuite) using the application in the [examples/autobahn
 subdirectory](https://github.com/gorilla/websocket/tree/master/examples/autobahn).
 
-### Gorilla WebSocket compared with other packages
-
-<table>
-<tr>
-<th></th>
-<th><a href="http://godoc.org/github.com/gorilla/websocket">github.com/gorilla</a></th>
-<th><a href="http://godoc.org/golang.org/x/net/websocket">golang.org/x/net</a></th>
-</tr>
-<tr>
-<tr><td colspan="3"><a href="http://tools.ietf.org/html/rfc6455">RFC 6455</a> Features</td></tr>
-<tr><td>Passes <a href="https://github.com/crossbario/autobahn-testsuite">Autobahn Test Suite</a></td><td><a href="https://github.com/gorilla/websocket/tree/master/examples/autobahn">Yes</a></td><td>No</td></tr>
-<tr><td>Receive <a href="https://tools.ietf.org/html/rfc6455#section-5.4">fragmented</a> message<td>Yes</td><td><a href="https://code.google.com/p/go/issues/detail?id=7632">No</a>, see note 1</td></tr>
-<tr><td>Send <a href="https://tools.ietf.org/html/rfc6455#section-5.5.1">close</a> message</td><td><a href="http://godoc.org/github.com/gorilla/websocket#hdr-Control_Messages">Yes</a></td><td><a href="https://code.google.com/p/go/issues/detail?id=4588">No</a></td></tr>
-<tr><td>Send <a href="https://tools.ietf.org/html/rfc6455#section-5.5.2">pings</a> and receive <a href="https://tools.ietf.org/html/rfc6455#section-5.5.3">pongs</a></td><td><a href="http://godoc.org/github.com/gorilla/websocket#hdr-Control_Messages">Yes</a></td><td>No</td></tr>
-<tr><td>Get the <a href="https://tools.ietf.org/html/rfc6455#section-5.6">type</a> of a received data message</td><td>Yes</td><td>Yes, see note 2</td></tr>
-<tr><td colspan="3">Other Features</tr></td>
-<tr><td><a href="https://tools.ietf.org/html/rfc7692">Compression Extensions</a></td><td>Experimental</td><td>No</td></tr>
-<tr><td>Read message using io.Reader</td><td><a href="http://godoc.org/github.com/gorilla/websocket#Conn.NextReader">Yes</a></td><td>No, see note 3</td></tr>
-<tr><td>Write message using io.WriteCloser</td><td><a href="http://godoc.org/github.com/gorilla/websocket#Conn.NextWriter">Yes</a></td><td>No, see note 3</td></tr>
-</table>
-
-Notes:
-
-1. Large messages are fragmented in [Chrome's new WebSocket implementation](http://www.ietf.org/mail-archive/web/hybi/current/msg10503.html).
-2. The application can get the type of a received data message by implementing
-   a [Codec marshal](http://godoc.org/golang.org/x/net/websocket#Codec.Marshal)
-   function.
-3. The go.net io.Reader and io.Writer operate across WebSocket frame boundaries.
-  Read returns when the input buffer is full or a frame boundary is
-  encountered. Each call to Write sends a single frame message. The Gorilla
-  io.Reader and io.WriteCloser operate on a single WebSocket message.
-

vendor/github.com/gorilla/websocket/client.go

@@ -48,15 +48,23 @@ func NewClient(netConn net.Conn, u *url.URL, requestHeader http.Header, readBufS
 }
 
 // A Dialer contains options for connecting to WebSocket server.
+//
+// It is safe to call Dialer's methods concurrently.
 type Dialer struct {
 	// NetDial specifies the dial function for creating TCP connections. If
 	// NetDial is nil, net.Dial is used.
 	NetDial func(network, addr string) (net.Conn, error)
 
 	// NetDialContext specifies the dial function for creating TCP connections. If
-	// NetDialContext is nil, net.DialContext is used.
+	// NetDialContext is nil, NetDial is used.
 	NetDialContext func(ctx context.Context, network, addr string) (net.Conn, error)
 
+	// NetDialTLSContext specifies the dial function for creating TLS/TCP connections. If
+	// NetDialTLSContext is nil, NetDialContext is used.
+	// If NetDialTLSContext is set, Dial assumes the TLS handshake is done there and
+	// TLSClientConfig is ignored.
+	NetDialTLSContext func(ctx context.Context, network, addr string) (net.Conn, error)
+
 	// Proxy specifies a function to return a proxy for a given
 	// Request. If the function returns a non-nil error, the
 	// request is aborted with the provided error.
@@ -65,6 +73,8 @@ type Dialer struct {
 
 	// TLSClientConfig specifies the TLS configuration to use with tls.Client.
 	// If nil, the default configuration is used.
+	// If either NetDialTLS or NetDialTLSContext are set, Dial assumes the TLS handshake
+	// is done there and TLSClientConfig is ignored.
 	TLSClientConfig *tls.Config
 
 	// HandshakeTimeout specifies the duration for the handshake to complete.
@@ -176,7 +186,7 @@ func (d *Dialer) DialContext(ctx context.Context, urlStr string, requestHeader h
 	}
 
 	req := &http.Request{
-		Method:     "GET",
+		Method:     http.MethodGet,
 		URL:        u,
 		Proto:      "HTTP/1.1",
 		ProtoMajor: 1,
@@ -237,13 +247,32 @@ func (d *Dialer) DialContext(ctx context.Context, urlStr string, requestHeader h
 	// Get network dial function.
 	var netDial func(network, add string) (net.Conn, error)
 
-	if d.NetDialContext != nil {
-		netDial = func(network, addr string) (net.Conn, error) {
-			return d.NetDialContext(ctx, network, addr)
+	switch u.Scheme {
+	case "http":
+		if d.NetDialContext != nil {
+			netDial = func(network, addr string) (net.Conn, error) {
+				return d.NetDialContext(ctx, network, addr)
+			}
+		} else if d.NetDial != nil {
+			netDial = d.NetDial
 		}
-	} else if d.NetDial != nil {
-		netDial = d.NetDial
-	} else {
+	case "https":
+		if d.NetDialTLSContext != nil {
+			netDial = func(network, addr string) (net.Conn, error) {
+				return d.NetDialTLSContext(ctx, network, addr)
+			}
+		} else if d.NetDialContext != nil {
+			netDial = func(network, addr string) (net.Conn, error) {
+				return d.NetDialContext(ctx, network, addr)
+			}
+		} else if d.NetDial != nil {
+			netDial = d.NetDial
+		}
+	default:
+		return nil, nil, errMalformedURL
+	}
+
+	if netDial == nil {
 		netDialer := &net.Dialer{}
 		netDial = func(network, addr string) (net.Conn, error) {
 			return netDialer.DialContext(ctx, network, addr)
@@ -304,7 +333,9 @@ func (d *Dialer) DialContext(ctx context.Context, urlStr string, requestHeader h
 		}
 	}()
 
-	if u.Scheme == "https" {
+	if u.Scheme == "https" && d.NetDialTLSContext == nil {
+		// If NetDialTLSContext is set, assume that the TLS handshake has already been done
+
 		cfg := cloneTLSConfig(d.TLSClientConfig)
 		if cfg.ServerName == "" {
 			cfg.ServerName = hostNoPort
@@ -312,11 +343,12 @@ func (d *Dialer) DialContext(ctx context.Context, urlStr string, requestHeader h
 		tlsConn := tls.Client(netConn, cfg)
 		netConn = tlsConn
 
-		var err error
-		if trace != nil {
-			err = doHandshakeWithTrace(trace, tlsConn, cfg)
-		} else {
-			err = doHandshake(tlsConn, cfg)
+		if trace != nil && trace.TLSHandshakeStart != nil {
+			trace.TLSHandshakeStart()
+		}
+		err := doHandshake(ctx, tlsConn, cfg)
+		if trace != nil && trace.TLSHandshakeDone != nil {
+			trace.TLSHandshakeDone(tlsConn.ConnectionState(), err)
 		}
 
 		if err != nil {
@@ -348,8 +380,8 @@ func (d *Dialer) DialContext(ctx context.Context, urlStr string, requestHeader h
 	}
 
 	if resp.StatusCode != 101 ||
-		!strings.EqualFold(resp.Header.Get("Upgrade"), "websocket") ||
-		!strings.EqualFold(resp.Header.Get("Connection"), "upgrade") ||
+		!tokenListContainsValue(resp.Header, "Upgrade", "websocket") ||
+		!tokenListContainsValue(resp.Header, "Connection", "upgrade") ||
 		resp.Header.Get("Sec-Websocket-Accept") != computeAcceptKey(challengeKey) {
 		// Before closing the network connection on return from this
 		// function, slurp up some of the response to aid application
@@ -382,14 +414,9 @@ func (d *Dialer) DialContext(ctx context.Context, urlStr string, requestHeader h
 	return conn, resp, nil
 }
 
-func doHandshake(tlsConn *tls.Conn, cfg *tls.Config) error {
-	if err := tlsConn.Handshake(); err != nil {
-		return err
+func cloneTLSConfig(cfg *tls.Config) *tls.Config {
+	if cfg == nil {
+		return &tls.Config{}
 	}
-	if !cfg.InsecureSkipVerify {
-		if err := tlsConn.VerifyHostname(cfg.ServerName); err != nil {
-			return err
-		}
-	}
-	return nil
+	return cfg.Clone()
 }

vendor/github.com/gorilla/websocket/client_clone.go

@@ -1,16 +0,0 @@
-// Copyright 2013 The Gorilla WebSocket Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style
-// license that can be found in the LICENSE file.
-
-// +build go1.8
-
-package websocket
-
-import "crypto/tls"
-
-func cloneTLSConfig(cfg *tls.Config) *tls.Config {
-	if cfg == nil {
-		return &tls.Config{}
-	}
-	return cfg.Clone()
-}

vendor/github.com/gorilla/websocket/client_clone_legacy.go

@@ -1,38 +0,0 @@
-// Copyright 2013 The Gorilla WebSocket Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style
-// license that can be found in the LICENSE file.
-
-// +build !go1.8
-
-package websocket
-
-import "crypto/tls"
-
-// cloneTLSConfig clones all public fields except the fields
-// SessionTicketsDisabled and SessionTicketKey. This avoids copying the
-// sync.Mutex in the sync.Once and makes it safe to call cloneTLSConfig on a
-// config in active use.
-func cloneTLSConfig(cfg *tls.Config) *tls.Config {
-	if cfg == nil {
-		return &tls.Config{}
-	}
-	return &tls.Config{
-		Rand:                     cfg.Rand,
-		Time:                     cfg.Time,
-		Certificates:             cfg.Certificates,
-		NameToCertificate:        cfg.NameToCertificate,
-		GetCertificate:           cfg.GetCertificate,
-		RootCAs:                  cfg.RootCAs,
-		NextProtos:               cfg.NextProtos,
-		ServerName:               cfg.ServerName,
-		ClientAuth:               cfg.ClientAuth,
-		ClientCAs:                cfg.ClientCAs,
-		InsecureSkipVerify:       cfg.InsecureSkipVerify,
-		CipherSuites:             cfg.CipherSuites,
-		PreferServerCipherSuites: cfg.PreferServerCipherSuites,
-		ClientSessionCache:       cfg.ClientSessionCache,
-		MinVersion:               cfg.MinVersion,
-		MaxVersion:               cfg.MaxVersion,
-		CurvePreferences:         cfg.CurvePreferences,
-	}
-}

vendor/github.com/gorilla/websocket/conn.go

@@ -13,6 +13,7 @@ import (
 	"math/rand"
 	"net"
 	"strconv"
+	"strings"
 	"sync"
 	"time"
 	"unicode/utf8"
@@ -401,6 +402,12 @@ func (c *Conn) write(frameType int, deadline time.Time, buf0, buf1 []byte) error
 	return nil
 }
 
+func (c *Conn) writeBufs(bufs ...[]byte) error {
+	b := net.Buffers(bufs)
+	_, err := b.WriteTo(c.conn)
+	return err
+}
+
 // WriteControl writes a control message with the given deadline. The allowed
 // message types are CloseMessage, PingMessage and PongMessage.
 func (c *Conn) WriteControl(messageType int, data []byte, deadline time.Time) error {
@@ -794,47 +801,69 @@ func (c *Conn) advanceFrame() (int, error) {
 	}
 
 	// 2. Read and parse first two bytes of frame header.
+	// To aid debugging, collect and report all errors in the first two bytes
+	// of the header.
+
+	var errors []string
 
 	p, err := c.read(2)
 	if err != nil {
 		return noFrame, err
 	}
 
-	final := p[0]&finalBit != 0
 	frameType := int(p[0] & 0xf)
+	final := p[0]&finalBit != 0
+	rsv1 := p[0]&rsv1Bit != 0
+	rsv2 := p[0]&rsv2Bit != 0
+	rsv3 := p[0]&rsv3Bit != 0
 	mask := p[1]&maskBit != 0
 	c.setReadRemaining(int64(p[1] & 0x7f))
 
 	c.readDecompress = false
-	if c.newDecompressionReader != nil && (p[0]&rsv1Bit) != 0 {
-		c.readDecompress = true
-		p[0] &^= rsv1Bit
+	if rsv1 {
+		if c.newDecompressionReader != nil {
+			c.readDecompress = true
+		} else {
+			errors = append(errors, "RSV1 set")
+		}
 	}
 
-	if rsv := p[0] & (rsv1Bit | rsv2Bit | rsv3Bit); rsv != 0 {
-		return noFrame, c.handleProtocolError("unexpected reserved bits 0x" + strconv.FormatInt(int64(rsv), 16))
+	if rsv2 {
+		errors = append(errors, "RSV2 set")
+	}
+
+	if rsv3 {
+		errors = append(errors, "RSV3 set")
 	}
 
 	switch frameType {
 	case CloseMessage, PingMessage, PongMessage:
 		if c.readRemaining > maxControlFramePayloadSize {
-			return noFrame, c.handleProtocolError("control frame length > 125")
+			errors = append(errors, "len > 125 for control")
 		}
 		if !final {
-			return noFrame, c.handleProtocolError("control frame not final")
+			errors = append(errors, "FIN not set on control")
 		}
 	case TextMessage, BinaryMessage:
 		if !c.readFinal {
-			return noFrame, c.handleProtocolError("message start before final message frame")
+			errors = append(errors, "data before FIN")
 		}
 		c.readFinal = final
 	case continuationFrame:
 		if c.readFinal {
-			return noFrame, c.handleProtocolError("continuation after final message frame")
+			errors = append(errors, "continuation after FIN")
 		}
 		c.readFinal = final
 	default:
-		return noFrame, c.handleProtocolError("unknown opcode " + strconv.Itoa(frameType))
+		errors = append(errors, "bad opcode "+strconv.Itoa(frameType))
+	}
+
+	if mask != c.isServer {
+		errors = append(errors, "bad MASK")
+	}
+
+	if len(errors) > 0 {
+		return noFrame, c.handleProtocolError(strings.Join(errors, ", "))
 	}
 
 	// 3. Read and parse frame length as per
@@ -872,10 +901,6 @@ func (c *Conn) advanceFrame() (int, error) {
 
 	// 4. Handle frame masking.
 
-	if mask != c.isServer {
-		return noFrame, c.handleProtocolError("incorrect mask flag")
-	}
-
 	if mask {
 		c.readMaskPos = 0
 		p, err := c.read(len(c.readMaskKey))
@@ -935,7 +960,7 @@ func (c *Conn) advanceFrame() (int, error) {
 		if len(payload) >= 2 {
 			closeCode = int(binary.BigEndian.Uint16(payload))
 			if !isValidReceivedCloseCode(closeCode) {
-				return noFrame, c.handleProtocolError("invalid close code")
+				return noFrame, c.handleProtocolError("bad close code " + strconv.Itoa(closeCode))
 			}
 			closeText = string(payload[2:])
 			if !utf8.ValidString(closeText) {
@@ -952,7 +977,11 @@ func (c *Conn) advanceFrame() (int, error) {
 }
 
 func (c *Conn) handleProtocolError(message string) error {
-	c.WriteControl(CloseMessage, FormatCloseMessage(CloseProtocolError, message), time.Now().Add(writeWait))
+	data := FormatCloseMessage(CloseProtocolError, message)
+	if len(data) > maxControlFramePayloadSize {
+		data = data[:maxControlFramePayloadSize]
+	}
+	c.WriteControl(CloseMessage, data, time.Now().Add(writeWait))
 	return errors.New("websocket: " + message)
 }
 

vendor/github.com/gorilla/websocket/conn_write.go

@@ -1,15 +0,0 @@
-// Copyright 2016 The Gorilla WebSocket Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style
-// license that can be found in the LICENSE file.
-
-// +build go1.8
-
-package websocket
-
-import "net"
-
-func (c *Conn) writeBufs(bufs ...[]byte) error {
-	b := net.Buffers(bufs)
-	_, err := b.WriteTo(c.conn)
-	return err
-}

vendor/github.com/gorilla/websocket/conn_write_legacy.go

@@ -1,18 +0,0 @@
-// Copyright 2016 The Gorilla WebSocket Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style
-// license that can be found in the LICENSE file.
-
-// +build !go1.8
-
-package websocket
-
-func (c *Conn) writeBufs(bufs ...[]byte) error {
-	for _, buf := range bufs {
-		if len(buf) > 0 {
-			if _, err := c.conn.Write(buf); err != nil {
-				return err
-			}
-		}
-	}
-	return nil
-}

vendor/github.com/gorilla/websocket/mask.go

@@ -2,6 +2,7 @@
 // this source code is governed by a BSD-style license that can be found in the
 // LICENSE file.
 
+//go:build !appengine
 // +build !appengine
 
 package websocket

vendor/github.com/gorilla/websocket/mask_safe.go

@@ -2,6 +2,7 @@
 // this source code is governed by a BSD-style license that can be found in the
 // LICENSE file.
 
+//go:build appengine
 // +build appengine
 
 package websocket

vendor/github.com/gorilla/websocket/proxy.go

@@ -48,7 +48,7 @@ func (hpd *httpProxyDialer) Dial(network string, addr string) (net.Conn, error)
 	}
 
 	connectReq := &http.Request{
-		Method: "CONNECT",
+		Method: http.MethodConnect,
 		URL:    &url.URL{Opaque: addr},
 		Host:   addr,
 		Header: connectHeader,

vendor/github.com/gorilla/websocket/server.go

@@ -23,6 +23,8 @@ func (e HandshakeError) Error() string { return e.message }
 
 // Upgrader specifies parameters for upgrading an HTTP connection to a
 // WebSocket connection.
+//
+// It is safe to call Upgrader's methods concurrently.
 type Upgrader struct {
 	// HandshakeTimeout specifies the duration for the handshake to complete.
 	HandshakeTimeout time.Duration
@@ -115,8 +117,8 @@ func (u *Upgrader) selectSubprotocol(r *http.Request, responseHeader http.Header
 // Upgrade upgrades the HTTP server connection to the WebSocket protocol.
 //
 // The responseHeader is included in the response to the client's upgrade
-// request. Use the responseHeader to specify cookies (Set-Cookie) and the
-// application negotiated subprotocol (Sec-WebSocket-Protocol).
+// request. Use the responseHeader to specify cookies (Set-Cookie). To specify
+// subprotocols supported by the server, set Upgrader.Subprotocols directly.
 //
 // If the upgrade fails, then Upgrade replies to the client with an HTTP error
 // response.
@@ -131,7 +133,7 @@ func (u *Upgrader) Upgrade(w http.ResponseWriter, r *http.Request, responseHeade
 		return u.returnError(w, r, http.StatusBadRequest, badHandshake+"'websocket' token not found in 'Upgrade' header")
 	}
 
-	if r.Method != "GET" {
+	if r.Method != http.MethodGet {
 		return u.returnError(w, r, http.StatusMethodNotAllowed, badHandshake+"request method is not GET")
 	}
 

vendor/github.com/gorilla/websocket/tls_handshake.go

@@ -0,0 +1,21 @@
+//go:build go1.17
+// +build go1.17
+
+package websocket
+
+import (
+	"context"
+	"crypto/tls"
+)
+
+func doHandshake(ctx context.Context, tlsConn *tls.Conn, cfg *tls.Config) error {
+	if err := tlsConn.HandshakeContext(ctx); err != nil {
+		return err
+	}
+	if !cfg.InsecureSkipVerify {
+		if err := tlsConn.VerifyHostname(cfg.ServerName); err != nil {
+			return err
+		}
+	}
+	return nil
+}

vendor/github.com/gorilla/websocket/tls_handshake_116.go

@@ -0,0 +1,21 @@
+//go:build !go1.17
+// +build !go1.17
+
+package websocket
+
+import (
+	"context"
+	"crypto/tls"
+)
+
+func doHandshake(ctx context.Context, tlsConn *tls.Conn, cfg *tls.Config) error {
+	if err := tlsConn.Handshake(); err != nil {
+		return err
+	}
+	if !cfg.InsecureSkipVerify {
+		if err := tlsConn.VerifyHostname(cfg.ServerName); err != nil {
+			return err
+		}
+	}
+	return nil
+}

vendor/github.com/gorilla/websocket/trace.go

@@ -1,19 +0,0 @@
-// +build go1.8
-
-package websocket
-
-import (
-	"crypto/tls"
-	"net/http/httptrace"
-)
-
-func doHandshakeWithTrace(trace *httptrace.ClientTrace, tlsConn *tls.Conn, cfg *tls.Config) error {
-	if trace.TLSHandshakeStart != nil {
-		trace.TLSHandshakeStart()
-	}
-	err := doHandshake(tlsConn, cfg)
-	if trace.TLSHandshakeDone != nil {
-		trace.TLSHandshakeDone(tlsConn.ConnectionState(), err)
-	}
-	return err
-}

vendor/github.com/gorilla/websocket/trace_17.go

@@ -1,12 +0,0 @@
-// +build !go1.8
-
-package websocket
-
-import (
-	"crypto/tls"
-	"net/http/httptrace"
-)
-
-func doHandshakeWithTrace(trace *httptrace.ClientTrace, tlsConn *tls.Conn, cfg *tls.Config) error {
-	return doHandshake(tlsConn, cfg)
-}

vendor/github.com/klauspost/compress/README.md

@@ -17,6 +17,19 @@ This package provides various compression algorithms.
 
 # changelog
 
+* Feb 17, 2022 (v1.14.3)
+	* flate: Improve fastest levels compression speed ~10% more throughput. [#482](https://github.com/klauspost/compress/pull/482) [#489](https://github.com/klauspost/compress/pull/489) [#490](https://github.com/klauspost/compress/pull/490) [#491](https://github.com/klauspost/compress/pull/491) [#494](https://github.com/klauspost/compress/pull/494)  [#478](https://github.com/klauspost/compress/pull/478)
+	* flate: Faster decompression speed, ~5-10%. [#483](https://github.com/klauspost/compress/pull/483)
+	* s2: Faster compression with Go v1.18 and amd64 microarch level 3+. [#484](https://github.com/klauspost/compress/pull/484) [#486](https://github.com/klauspost/compress/pull/486)
+
+* Jan 25, 2022 (v1.14.2)
+	* zstd: improve header decoder by @dsnet  [#476](https://github.com/klauspost/compress/pull/476)
+	* zstd: Add bigger default blocks  [#469](https://github.com/klauspost/compress/pull/469)
+	* zstd: Remove unused decompression buffer [#470](https://github.com/klauspost/compress/pull/470)
+	* zstd: Fix logically dead code by @ningmingxiao [#472](https://github.com/klauspost/compress/pull/472)
+	* flate: Improve level 7-9 [#471](https://github.com/klauspost/compress/pull/471) [#473](https://github.com/klauspost/compress/pull/473)
+	* zstd: Add noasm tag for xxhash [#475](https://github.com/klauspost/compress/pull/475)
+
 * Jan 11, 2022 (v1.14.1)
 	* s2: Add stream index in [#462](https://github.com/klauspost/compress/pull/462)
 	* flate: Speed and efficiency improvements in [#439](https://github.com/klauspost/compress/pull/439) [#461](https://github.com/klauspost/compress/pull/461) [#455](https://github.com/klauspost/compress/pull/455) [#452](https://github.com/klauspost/compress/pull/452) [#458](https://github.com/klauspost/compress/pull/458)
@@ -53,6 +66,9 @@ This package provides various compression algorithms.
 	* zstd: Detect short invalid signatures [#382](https://github.com/klauspost/compress/pull/382)
 	* zstd: Spawn decoder goroutine only if needed. [#380](https://github.com/klauspost/compress/pull/380)
 
+<details>
+	<summary>See changes to v1.12.x</summary>
+	
 * May 25, 2021 (v1.12.3)
 	* deflate: Better/faster Huffman encoding [#374](https://github.com/klauspost/compress/pull/374)
 	* deflate: Allocate less for history. [#375](https://github.com/klauspost/compress/pull/375)
@@ -74,9 +90,10 @@ This package provides various compression algorithms.
 	* s2c/s2d/s2sx: Always truncate when writing files [#352](https://github.com/klauspost/compress/pull/352)
 	* zstd: Reduce memory usage further when using [WithLowerEncoderMem](https://pkg.go.dev/github.com/klauspost/compress/zstd#WithLowerEncoderMem) [#346](https://github.com/klauspost/compress/pull/346)
 	* s2: Fix potential problem with amd64 assembly and profilers [#349](https://github.com/klauspost/compress/pull/349)
+</details>
 
 <details>
-	<summary>See changes prior to v1.12.1</summary>
+	<summary>See changes to v1.11.x</summary>
 	
 * Mar 26, 2021 (v1.11.13)
 	* zstd: Big speedup on small dictionary encodes [#344](https://github.com/klauspost/compress/pull/344) [#345](https://github.com/klauspost/compress/pull/345)
@@ -135,7 +152,7 @@ This package provides various compression algorithms.
 </details>
 
 <details>
-	<summary>See changes prior to v1.11.0</summary>
+	<summary>See changes to v1.10.x</summary>
  
 * July 8, 2020 (v1.10.11) 
 	* zstd: Fix extra block when compressing with ReadFrom. [#278](https://github.com/klauspost/compress/pull/278)
@@ -297,11 +314,6 @@ This package provides various compression algorithms.
 
 # deflate usage
 
-* [High Throughput Benchmark](http://blog.klauspost.com/go-gzipdeflate-benchmarks/).
-* [Small Payload/Webserver Benchmarks](http://blog.klauspost.com/gzip-performance-for-go-webservers/).
-* [Linear Time Compression](http://blog.klauspost.com/constant-time-gzipzip-compression/).
-* [Re-balancing Deflate Compression Levels](https://blog.klauspost.com/rebalancing-deflate-compression-levels/)
-
 The packages are drop-in replacements for standard libraries. Simply replace the import path to use them:
 
 | old import         | new import                              | Documentation
@@ -323,6 +335,8 @@ Memory usage is typically 1MB for a Writer. stdlib is in the same range.
 If you expect to have a lot of concurrently allocated Writers consider using 
 the stateless compress described below.
 
+For compression performance, see: [this spreadsheet](https://docs.google.com/spreadsheets/d/1nuNE2nPfuINCZJRMt6wFWhKpToF95I47XjSsc-1rbPQ/edit?usp=sharing).
+
 # Stateless compression
 
 This package offers stateless compression as a special option for gzip/deflate. 

vendor/github.com/klauspost/compress/flate/fast_encoder.go

@@ -179,7 +179,7 @@ func (e *fastGen) matchlen(s, t int32, src []byte) int32 {
 // matchlenLong will return the match length between offsets and t in src.
 // It is assumed that s > t, that t >=0 and s < len(src).
 func (e *fastGen) matchlenLong(s, t int32, src []byte) int32 {
-	if debugDecode {
+	if debugDeflate {
 		if t >= s {
 			panic(fmt.Sprint("t >=s:", t, s))
 		}

vendor/github.com/klauspost/compress/flate/huffman_bit_writer.go

@@ -8,6 +8,7 @@ import (
 	"encoding/binary"
 	"fmt"
 	"io"
+	"math"
 )
 
 const (
@@ -24,6 +25,10 @@ const (
 	codegenCodeCount = 19
 	badCode          = 255
 
+	// maxPredefinedTokens is the maximum number of tokens
+	// where we check if fixed size is smaller.
+	maxPredefinedTokens = 250
+
 	// bufferFlushSize indicates the buffer size
 	// after which bytes are flushed to the writer.
 	// Should preferably be a multiple of 6, since
@@ -36,8 +41,11 @@ const (
 	bufferSize = bufferFlushSize + 8
 )
 
+// Minimum length code that emits bits.
+const lengthExtraBitsMinCode = 8
+
 // The number of extra bits needed by length code X - LENGTH_CODES_START.
-var lengthExtraBits = [32]int8{
+var lengthExtraBits = [32]uint8{
 	/* 257 */ 0, 0, 0,
 	/* 260 */ 0, 0, 0, 0, 0, 1, 1, 1, 1, 2,
 	/* 270 */ 2, 2, 2, 3, 3, 3, 3, 4, 4, 4,
@@ -51,6 +59,9 @@ var lengthBase = [32]uint8{
 	64, 80, 96, 112, 128, 160, 192, 224, 255,
 }
 
+// Minimum offset code that emits bits.
+const offsetExtraBitsMinCode = 4
+
 // offset code word extra bits.
 var offsetExtraBits = [32]int8{
 	0, 0, 0, 0, 1, 1, 2, 2, 3, 3,
@@ -78,10 +89,10 @@ func init() {
 
 	for i := range offsetCombined[:] {
 		// Don't use extended window values...
-		if offsetBase[i] > 0x006000 {
+		if offsetExtraBits[i] == 0 || offsetBase[i] > 0x006000 {
 			continue
 		}
-		offsetCombined[i] = uint32(offsetExtraBits[i])<<16 | (offsetBase[i])
+		offsetCombined[i] = uint32(offsetExtraBits[i]) | (offsetBase[i] << 8)
 	}
 }
 
@@ -97,7 +108,7 @@ type huffmanBitWriter struct {
 	// Data waiting to be written is bytes[0:nbytes]
 	// and then the low nbits of bits.
 	bits            uint64
-	nbits           uint16
+	nbits           uint8
 	nbytes          uint8
 	lastHuffMan     bool
 	literalEncoding *huffmanEncoder
@@ -215,7 +226,7 @@ func (w *huffmanBitWriter) write(b []byte) {
 	_, w.err = w.writer.Write(b)
 }
 
-func (w *huffmanBitWriter) writeBits(b int32, nb uint16) {
+func (w *huffmanBitWriter) writeBits(b int32, nb uint8) {
 	w.bits |= uint64(b) << (w.nbits & 63)
 	w.nbits += nb
 	if w.nbits >= 48 {
@@ -571,7 +582,10 @@ func (w *huffmanBitWriter) writeBlock(tokens *tokens, eof bool, input []byte) {
 	// Fixed Huffman baseline.
 	var literalEncoding = fixedLiteralEncoding
 	var offsetEncoding = fixedOffsetEncoding
-	var size = w.fixedSize(extraBits)
+	var size = math.MaxInt32
+	if tokens.n < maxPredefinedTokens {
+		size = w.fixedSize(extraBits)
+	}
 
 	// Dynamic Huffman?
 	var numCodegens int
@@ -672,19 +686,21 @@ func (w *huffmanBitWriter) writeBlockDynamic(tokens *tokens, eof bool, input []b
 			size = reuseSize
 		}
 
-		if preSize := w.fixedSize(extraBits) + 7; usePrefs && preSize < size {
-			// Check if we get a reasonable size decrease.
-			if storable && ssize <= size {
-				w.writeStoredHeader(len(input), eof)
-				w.writeBytes(input)
+		if tokens.n < maxPredefinedTokens {
+			if preSize := w.fixedSize(extraBits) + 7; usePrefs && preSize < size {
+				// Check if we get a reasonable size decrease.
+				if storable && ssize <= size {
+					w.writeStoredHeader(len(input), eof)
+					w.writeBytes(input)
+					return
+				}
+				w.writeFixedHeader(eof)
+				if !sync {
+					tokens.AddEOB()
+				}
+				w.writeTokens(tokens.Slice(), fixedLiteralEncoding.codes, fixedOffsetEncoding.codes)
 				return
 			}
-			w.writeFixedHeader(eof)
-			if !sync {
-				tokens.AddEOB()
-			}
-			w.writeTokens(tokens.Slice(), fixedLiteralEncoding.codes, fixedOffsetEncoding.codes)
-			return
 		}
 		// Check if we get a reasonable size decrease.
 		if storable && ssize <= size {
@@ -717,19 +733,21 @@ func (w *huffmanBitWriter) writeBlockDynamic(tokens *tokens, eof bool, input []b
 		size, numCodegens = w.dynamicSize(w.literalEncoding, w.offsetEncoding, extraBits)
 
 		// Store predefined, if we don't get a reasonable improvement.
-		if preSize := w.fixedSize(extraBits); usePrefs && preSize <= size {
-			// Store bytes, if we don't get an improvement.
-			if storable && ssize <= preSize {
-				w.writeStoredHeader(len(input), eof)
-				w.writeBytes(input)
+		if tokens.n < maxPredefinedTokens {
+			if preSize := w.fixedSize(extraBits); usePrefs && preSize <= size {
+				// Store bytes, if we don't get an improvement.
+				if storable && ssize <= preSize {
+					w.writeStoredHeader(len(input), eof)
+					w.writeBytes(input)
+					return
+				}
+				w.writeFixedHeader(eof)
+				if !sync {
+					tokens.AddEOB()
+				}
+				w.writeTokens(tokens.Slice(), fixedLiteralEncoding.codes, fixedOffsetEncoding.codes)
 				return
 			}
-			w.writeFixedHeader(eof)
-			if !sync {
-				tokens.AddEOB()
-			}
-			w.writeTokens(tokens.Slice(), fixedLiteralEncoding.codes, fixedOffsetEncoding.codes)
-			return
 		}
 
 		if storable && ssize <= size {
@@ -833,9 +851,9 @@ func (w *huffmanBitWriter) writeTokens(tokens []token, leCodes, oeCodes []hcode)
 	bits, nbits, nbytes := w.bits, w.nbits, w.nbytes
 
 	for _, t := range tokens {
-		if t < matchType {
+		if t < 256 {
 			//w.writeCode(lits[t.literal()])
-			c := lits[t.literal()]
+			c := lits[t]
 			bits |= uint64(c.code) << (nbits & 63)
 			nbits += c.len
 			if nbits >= 48 {
@@ -858,12 +876,12 @@ func (w *huffmanBitWriter) writeTokens(tokens []token, leCodes, oeCodes []hcode)
 
 		// Write the length
 		length := t.length()
-		lengthCode := lengthCode(length)
+		lengthCode := lengthCode(length) & 31
 		if false {
-			w.writeCode(lengths[lengthCode&31])
+			w.writeCode(lengths[lengthCode])
 		} else {
 			// inlined
-			c := lengths[lengthCode&31]
+			c := lengths[lengthCode]
 			bits |= uint64(c.code) << (nbits & 63)
 			nbits += c.len
 			if nbits >= 48 {
@@ -883,10 +901,10 @@ func (w *huffmanBitWriter) writeTokens(tokens []token, leCodes, oeCodes []hcode)
 			}
 		}
 
-		extraLengthBits := uint16(lengthExtraBits[lengthCode&31])
-		if extraLengthBits > 0 {
+		if lengthCode >= lengthExtraBitsMinCode {
+			extraLengthBits := lengthExtraBits[lengthCode]
 			//w.writeBits(extraLength, extraLengthBits)
-			extraLength := int32(length - lengthBase[lengthCode&31])
+			extraLength := int32(length - lengthBase[lengthCode])
 			bits |= uint64(extraLength) << (nbits & 63)
 			nbits += extraLengthBits
 			if nbits >= 48 {
@@ -907,10 +925,9 @@ func (w *huffmanBitWriter) writeTokens(tokens []token, leCodes, oeCodes []hcode)
 		}
 		// Write the offset
 		offset := t.offset()
-		offsetCode := offset >> 16
-		offset &= matchOffsetOnlyMask
+		offsetCode := (offset >> 16) & 31
 		if false {
-			w.writeCode(offs[offsetCode&31])
+			w.writeCode(offs[offsetCode])
 		} else {
 			// inlined
 			c := offs[offsetCode]
@@ -932,11 +949,12 @@ func (w *huffmanBitWriter) writeTokens(tokens []token, leCodes, oeCodes []hcode)
 				}
 			}
 		}
-		offsetComb := offsetCombined[offsetCode]
-		if offsetComb > 1<<16 {
+
+		if offsetCode >= offsetExtraBitsMinCode {
+			offsetComb := offsetCombined[offsetCode]
 			//w.writeBits(extraOffset, extraOffsetBits)
-			bits |= uint64(offset-(offsetComb&0xffff)) << (nbits & 63)
-			nbits += uint16(offsetComb >> 16)
+			bits |= uint64((offset-(offsetComb>>8))&matchOffsetOnlyMask) << (nbits & 63)
+			nbits += uint8(offsetComb)
 			if nbits >= 48 {
 				binary.LittleEndian.PutUint64(w.bytes[nbytes:], bits)
 				//*(*uint64)(unsafe.Pointer(&w.bytes[nbytes])) = bits
@@ -1002,6 +1020,29 @@ func (w *huffmanBitWriter) writeBlockHuff(eof bool, input []byte, sync bool) {
 	// https://stackoverflow.com/a/25454430
 	const guessHeaderSizeBits = 70 * 8
 	histogram(input, w.literalFreq[:numLiterals], fill)
+	ssize, storable := w.storedSize(input)
+	if storable && len(input) > 1024 {
+		// Quick check for incompressible content.
+		abs := float64(0)
+		avg := float64(len(input)) / 256
+		max := float64(len(input) * 2)
+		for _, v := range w.literalFreq[:256] {
+			diff := float64(v) - avg
+			abs += diff * diff
+			if abs > max {
+				break
+			}
+		}
+		if abs < max {
+			if debugDeflate {
+				fmt.Println("stored", abs, "<", max)
+			}
+			// No chance we can compress this...
+			w.writeStoredHeader(len(input), eof)
+			w.writeBytes(input)
+			return
+		}
+	}
 	w.literalFreq[endBlockMarker] = 1
 	w.tmpLitEncoding.generate(w.literalFreq[:numLiterals], 15)
 	if fill {
@@ -1019,8 +1060,10 @@ func (w *huffmanBitWriter) writeBlockHuff(eof bool, input []byte, sync bool) {
 	estBits += estBits >> w.logNewTablePenalty
 
 	// Store bytes, if we don't get a reasonable improvement.
-	ssize, storable := w.storedSize(input)
 	if storable && ssize <= estBits {
+		if debugDeflate {
+			fmt.Println("stored,", ssize, "<=", estBits)
+		}
 		w.writeStoredHeader(len(input), eof)
 		w.writeBytes(input)
 		return
@@ -1031,7 +1074,7 @@ func (w *huffmanBitWriter) writeBlockHuff(eof bool, input []byte, sync bool) {
 
 		if estBits < reuseSize {
 			if debugDeflate {
-				//fmt.Println("not reusing, reuse:", reuseSize/8, "> new:", estBits/8, "- header est:", w.lastHeader/8)
+				fmt.Println("NOT reusing, reuse:", reuseSize/8, "> new:", estBits/8, "header est:", w.lastHeader/8, "bytes")
 			}
 			// We owe an EOB
 			w.writeCode(w.literalEncoding.codes[endBlockMarker])
@@ -1065,6 +1108,9 @@ func (w *huffmanBitWriter) writeBlockHuff(eof bool, input []byte, sync bool) {
 	// Go 1.16 LOVES having these on stack. At least 1.5x the speed.
 	bits, nbits, nbytes := w.bits, w.nbits, w.nbytes
 
+	if debugDeflate {
+		count -= int(nbytes)*8 + int(nbits)
+	}
 	// Unroll, write 3 codes/loop.
 	// Fastest number of unrolls.
 	for len(input) > 3 {
@@ -1074,13 +1120,16 @@ func (w *huffmanBitWriter) writeBlockHuff(eof bool, input []byte, sync bool) {
 			binary.LittleEndian.PutUint64(w.bytes[nbytes:], bits)
 			bits >>= (n * 8) & 63
 			nbits -= n * 8
-			nbytes += uint8(n)
+			nbytes += n
 		}
 		if nbytes >= bufferFlushSize {
 			if w.err != nil {
 				nbytes = 0
 				return
 			}
+			if debugDeflate {
+				count += int(nbytes) * 8
+			}
 			_, w.err = w.writer.Write(w.bytes[:nbytes])
 			nbytes = 0
 		}
@@ -1096,13 +1145,6 @@ func (w *huffmanBitWriter) writeBlockHuff(eof bool, input []byte, sync bool) {
 
 	// Remaining...
 	for _, t := range input {
-		// Bitwriting inlined, ~30% speedup
-		c := encoding[t]
-		bits |= uint64(c.code) << (nbits & 63)
-		nbits += c.len
-		if debugDeflate {
-			count += int(c.len)
-		}
 		if nbits >= 48 {
 			binary.LittleEndian.PutUint64(w.bytes[nbytes:], bits)
 			//*(*uint64)(unsafe.Pointer(&w.bytes[nbytes])) = bits
@@ -1114,17 +1156,33 @@ func (w *huffmanBitWriter) writeBlockHuff(eof bool, input []byte, sync bool) {
 					nbytes = 0
 					return
 				}
+				if debugDeflate {
+					count += int(nbytes) * 8
+				}
 				_, w.err = w.writer.Write(w.bytes[:nbytes])
 				nbytes = 0
 			}
 		}
+		// Bitwriting inlined, ~30% speedup
+		c := encoding[t]
+		bits |= uint64(c.code) << (nbits & 63)
+		nbits += c.len
+		if debugDeflate {
+			count += int(c.len)
+		}
 	}
 	// Restore...
 	w.bits, w.nbits, w.nbytes = bits, nbits, nbytes
 
 	if debugDeflate {
-		fmt.Println("wrote", count/8, "bytes")
+		nb := count + int(nbytes)*8 + int(nbits)
+		fmt.Println("wrote", nb, "bits,", nb/8, "bytes.")
 	}
+	// Flush if needed to have space.
+	if w.nbits >= 48 {
+		w.writeOutBits()
+	}
+
 	if eof || sync {
 		w.writeCode(w.literalEncoding.codes[endBlockMarker])
 		w.lastHeader = 0

vendor/github.com/klauspost/compress/flate/huffman_code.go

@@ -17,7 +17,8 @@ const (
 
 // hcode is a huffman code with a bit code and bit length.
 type hcode struct {
-	code, len uint16
+	code uint16
+	len  uint8
 }
 
 type huffmanEncoder struct {
@@ -56,7 +57,7 @@ type levelInfo struct {
 }
 
 // set sets the code and length of an hcode.
-func (h *hcode) set(code uint16, length uint16) {
+func (h *hcode) set(code uint16, length uint8) {
 	h.len = length
 	h.code = code
 }
@@ -80,7 +81,7 @@ func generateFixedLiteralEncoding() *huffmanEncoder {
 	var ch uint16
 	for ch = 0; ch < literalCount; ch++ {
 		var bits uint16
-		var size uint16
+		var size uint8
 		switch {
 		case ch < 144:
 			// size 8, 000110000  .. 10111111
@@ -99,7 +100,7 @@ func generateFixedLiteralEncoding() *huffmanEncoder {
 			bits = ch + 192 - 280
 			size = 8
 		}
-		codes[ch] = hcode{code: reverseBits(bits, byte(size)), len: size}
+		codes[ch] = hcode{code: reverseBits(bits, size), len: size}
 	}
 	return h
 }
@@ -187,14 +188,19 @@ func (h *huffmanEncoder) bitCounts(list []literalNode, maxBits int32) []int32 {
 	// of the level j ancestor.
 	var leafCounts [maxBitsLimit][maxBitsLimit]int32
 
+	// Descending to only have 1 bounds check.
+	l2f := int32(list[2].freq)
+	l1f := int32(list[1].freq)
+	l0f := int32(list[0].freq) + int32(list[1].freq)
+
 	for level := int32(1); level <= maxBits; level++ {
 		// For every level, the first two items are the first two characters.
 		// We initialize the levels as if we had already figured this out.
 		levels[level] = levelInfo{
 			level:        level,
-			lastFreq:     int32(list[1].freq),
-			nextCharFreq: int32(list[2].freq),
-			nextPairFreq: int32(list[0].freq) + int32(list[1].freq),
+			lastFreq:     l1f,
+			nextCharFreq: l2f,
+			nextPairFreq: l0f,
 		}
 		leafCounts[level][level] = 2
 		if level == 1 {
@@ -205,8 +211,8 @@ func (h *huffmanEncoder) bitCounts(list []literalNode, maxBits int32) []int32 {
 	// We need a total of 2*n - 2 items at top level and have already generated 2.
 	levels[maxBits].needed = 2*n - 4
 
-	level := maxBits
-	for {
+	level := uint32(maxBits)
+	for level < 16 {
 		l := &levels[level]
 		if l.nextPairFreq == math.MaxInt32 && l.nextCharFreq == math.MaxInt32 {
 			// We've run out of both leafs and pairs.
@@ -238,7 +244,13 @@ func (h *huffmanEncoder) bitCounts(list []literalNode, maxBits int32) []int32 {
 			// more values in the level below
 			l.lastFreq = l.nextPairFreq
 			// Take leaf counts from the lower level, except counts[level] remains the same.
-			copy(leafCounts[level][:level], leafCounts[level-1][:level])
+			if true {
+				save := leafCounts[level][level]
+				leafCounts[level] = leafCounts[level-1]
+				leafCounts[level][level] = save
+			} else {
+				copy(leafCounts[level][:level], leafCounts[level-1][:level])
+			}
 			levels[l.level-1].needed = 2
 		}
 
@@ -296,7 +308,7 @@ func (h *huffmanEncoder) assignEncodingAndSize(bitCount []int32, list []literalN
 
 		sortByLiteral(chunk)
 		for _, node := range chunk {
-			h.codes[node.literal] = hcode{code: reverseBits(code, uint8(n)), len: uint16(n)}
+			h.codes[node.literal] = hcode{code: reverseBits(code, uint8(n)), len: uint8(n)}
 			code++
 		}
 		list = list[0 : len(list)-int(bits)]
@@ -309,6 +321,7 @@ func (h *huffmanEncoder) assignEncodingAndSize(bitCount []int32, list []literalN
 // maxBits  The maximum number of bits to use for any literal.
 func (h *huffmanEncoder) generate(freq []uint16, maxBits int32) {
 	list := h.freqcache[:len(freq)+1]
+	codes := h.codes[:len(freq)]
 	// Number of non-zero literals
 	count := 0
 	// Set list to be the set of all non-zero literals and their frequencies
@@ -317,11 +330,10 @@ func (h *huffmanEncoder) generate(freq []uint16, maxBits int32) {
 			list[count] = literalNode{uint16(i), f}
 			count++
 		} else {
-			list[count] = literalNode{}
-			h.codes[i].len = 0
+			codes[i].len = 0
 		}
 	}
-	list[len(freq)] = literalNode{}
+	list[count] = literalNode{}
 
 	list = list[:count]
 	if count <= 2 {

vendor/github.com/klauspost/compress/flate/inflate.go

@@ -36,6 +36,13 @@ type lengthExtra struct {
 
 var decCodeToLen = [32]lengthExtra{{length: 0x0, extra: 0x0}, {length: 0x1, extra: 0x0}, {length: 0x2, extra: 0x0}, {length: 0x3, extra: 0x0}, {length: 0x4, extra: 0x0}, {length: 0x5, extra: 0x0}, {length: 0x6, extra: 0x0}, {length: 0x7, extra: 0x0}, {length: 0x8, extra: 0x1}, {length: 0xa, extra: 0x1}, {length: 0xc, extra: 0x1}, {length: 0xe, extra: 0x1}, {length: 0x10, extra: 0x2}, {length: 0x14, extra: 0x2}, {length: 0x18, extra: 0x2}, {length: 0x1c, extra: 0x2}, {length: 0x20, extra: 0x3}, {length: 0x28, extra: 0x3}, {length: 0x30, extra: 0x3}, {length: 0x38, extra: 0x3}, {length: 0x40, extra: 0x4}, {length: 0x50, extra: 0x4}, {length: 0x60, extra: 0x4}, {length: 0x70, extra: 0x4}, {length: 0x80, extra: 0x5}, {length: 0xa0, extra: 0x5}, {length: 0xc0, extra: 0x5}, {length: 0xe0, extra: 0x5}, {length: 0xff, extra: 0x0}, {length: 0x0, extra: 0x0}, {length: 0x0, extra: 0x0}, {length: 0x0, extra: 0x0}}
 
+var bitMask32 = [32]uint32{
+	0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 0xFF,
+	0x1FF, 0x3FF, 0x7FF, 0xFFF, 0x1FFF, 0x3FFF, 0x7FFF, 0xFFFF,
+	0x1ffff, 0x3ffff, 0x7FFFF, 0xfFFFF, 0x1fFFFF, 0x3fFFFF, 0x7fFFFF, 0xffFFFF,
+	0x1ffFFFF, 0x3ffFFFF, 0x7ffFFFF, 0xfffFFFF, 0x1fffFFFF, 0x3fffFFFF, 0x7fffFFFF,
+} // up to 32 bits
+
 // Initialize the fixedHuffmanDecoder only once upon first use.
 var fixedOnce sync.Once
 var fixedHuffmanDecoder huffmanDecoder
@@ -559,221 +566,6 @@ func (f *decompressor) readHuffman() error {
 	return nil
 }
 
-// Decode a single Huffman block from f.
-// hl and hd are the Huffman states for the lit/length values
-// and the distance values, respectively. If hd == nil, using the
-// fixed distance encoding associated with fixed Huffman blocks.
-func (f *decompressor) huffmanBlockGeneric() {
-	const (
-		stateInit = iota // Zero value must be stateInit
-		stateDict
-	)
-
-	switch f.stepState {
-	case stateInit:
-		goto readLiteral
-	case stateDict:
-		goto copyHistory
-	}
-
-readLiteral:
-	// Read literal and/or (length, distance) according to RFC section 3.2.3.
-	{
-		var v int
-		{
-			// Inlined v, err := f.huffSym(f.hl)
-			// Since a huffmanDecoder can be empty or be composed of a degenerate tree
-			// with single element, huffSym must error on these two edge cases. In both
-			// cases, the chunks slice will be 0 for the invalid sequence, leading it
-			// satisfy the n == 0 check below.
-			n := uint(f.hl.maxRead)
-			// Optimization. Compiler isn't smart enough to keep f.b,f.nb in registers,
-			// but is smart enough to keep local variables in registers, so use nb and b,
-			// inline call to moreBits and reassign b,nb back to f on return.
-			nb, b := f.nb, f.b
-			for {
-				for nb < n {
-					c, err := f.r.ReadByte()
-					if err != nil {
-						f.b = b
-						f.nb = nb
-						f.err = noEOF(err)
-						return
-					}
-					f.roffset++
-					b |= uint32(c) << (nb & regSizeMaskUint32)
-					nb += 8
-				}
-				chunk := f.hl.chunks[b&(huffmanNumChunks-1)]
-				n = uint(chunk & huffmanCountMask)
-				if n > huffmanChunkBits {
-					chunk = f.hl.links[chunk>>huffmanValueShift][(b>>huffmanChunkBits)&f.hl.linkMask]
-					n = uint(chunk & huffmanCountMask)
-				}
-				if n <= nb {
-					if n == 0 {
-						f.b = b
-						f.nb = nb
-						if debugDecode {
-							fmt.Println("huffsym: n==0")
-						}
-						f.err = CorruptInputError(f.roffset)
-						return
-					}
-					f.b = b >> (n & regSizeMaskUint32)
-					f.nb = nb - n
-					v = int(chunk >> huffmanValueShift)
-					break
-				}
-			}
-		}
-
-		var n uint // number of bits extra
-		var length int
-		var err error
-		switch {
-		case v < 256:
-			f.dict.writeByte(byte(v))
-			if f.dict.availWrite() == 0 {
-				f.toRead = f.dict.readFlush()
-				f.step = (*decompressor).huffmanBlockGeneric
-				f.stepState = stateInit
-				return
-			}
-			goto readLiteral
-		case v == 256:
-			f.finishBlock()
-			return
-		// otherwise, reference to older data
-		case v < 265:
-			length = v - (257 - 3)
-			n = 0
-		case v < 269:
-			length = v*2 - (265*2 - 11)
-			n = 1
-		case v < 273:
-			length = v*4 - (269*4 - 19)
-			n = 2
-		case v < 277:
-			length = v*8 - (273*8 - 35)
-			n = 3
-		case v < 281:
-			length = v*16 - (277*16 - 67)
-			n = 4
-		case v < 285:
-			length = v*32 - (281*32 - 131)
-			n = 5
-		case v < maxNumLit:
-			length = 258
-			n = 0
-		default:
-			if debugDecode {
-				fmt.Println(v, ">= maxNumLit")
-			}
-			f.err = CorruptInputError(f.roffset)
-			return
-		}
-		if n > 0 {
-			for f.nb < n {
-				if err = f.moreBits(); err != nil {
-					if debugDecode {
-						fmt.Println("morebits n>0:", err)
-					}
-					f.err = err
-					return
-				}
-			}
-			length += int(f.b & uint32(1<<(n&regSizeMaskUint32)-1))
-			f.b >>= n & regSizeMaskUint32
-			f.nb -= n
-		}
-
-		var dist uint32
-		if f.hd == nil {
-			for f.nb < 5 {
-				if err = f.moreBits(); err != nil {
-					if debugDecode {
-						fmt.Println("morebits f.nb<5:", err)
-					}
-					f.err = err
-					return
-				}
-			}
-			dist = uint32(bits.Reverse8(uint8(f.b & 0x1F << 3)))
-			f.b >>= 5
-			f.nb -= 5
-		} else {
-			sym, err := f.huffSym(f.hd)
-			if err != nil {
-				if debugDecode {
-					fmt.Println("huffsym:", err)
-				}
-				f.err = err
-				return
-			}
-			dist = uint32(sym)
-		}
-
-		switch {
-		case dist < 4:
-			dist++
-		case dist < maxNumDist:
-			nb := uint(dist-2) >> 1
-			// have 1 bit in bottom of dist, need nb more.
-			extra := (dist & 1) << (nb & regSizeMaskUint32)
-			for f.nb < nb {
-				if err = f.moreBits(); err != nil {
-					if debugDecode {
-						fmt.Println("morebits f.nb<nb:", err)
-					}
-					f.err = err
-					return
-				}
-			}
-			extra |= f.b & uint32(1<<(nb&regSizeMaskUint32)-1)
-			f.b >>= nb & regSizeMaskUint32
-			f.nb -= nb
-			dist = 1<<((nb+1)&regSizeMaskUint32) + 1 + extra
-		default:
-			if debugDecode {
-				fmt.Println("dist too big:", dist, maxNumDist)
-			}
-			f.err = CorruptInputError(f.roffset)
-			return
-		}
-
-		// No check on length; encoding can be prescient.
-		if dist > uint32(f.dict.histSize()) {
-			if debugDecode {
-				fmt.Println("dist > f.dict.histSize():", dist, f.dict.histSize())
-			}
-			f.err = CorruptInputError(f.roffset)
-			return
-		}
-
-		f.copyLen, f.copyDist = length, int(dist)
-		goto copyHistory
-	}
-
-copyHistory:
-	// Perform a backwards copy according to RFC section 3.2.3.
-	{
-		cnt := f.dict.tryWriteCopy(f.copyDist, f.copyLen)
-		if cnt == 0 {
-			cnt = f.dict.writeCopy(f.copyDist, f.copyLen)
-		}
-		f.copyLen -= cnt
-
-		if f.dict.availWrite() == 0 || f.copyLen > 0 {
-			f.toRead = f.dict.readFlush()
-			f.step = (*decompressor).huffmanBlockGeneric // We need to continue this work
-			f.stepState = stateDict
-			return
-		}
-		goto readLiteral
-	}
-}
-
 // Copy a single uncompressed data block from input to output.
 func (f *decompressor) dataBlock() {
 	// Uncompressed.

vendor/github.com/klauspost/compress/flate/level1.go

@@ -1,6 +1,10 @@
 package flate
 
-import "fmt"
+import (
+	"encoding/binary"
+	"fmt"
+	"math/bits"
+)
 
 // fastGen maintains the table for matches,
 // and the previous byte block for level 2.
@@ -116,7 +120,32 @@ func (e *fastEncL1) Encode(dst *tokens, src []byte) {
 
 			// Extend the 4-byte match as long as possible.
 			t := candidate.offset - e.cur
-			l := e.matchlenLong(s+4, t+4, src) + 4
+			var l = int32(4)
+			if false {
+				l = e.matchlenLong(s+4, t+4, src) + 4
+			} else {
+				// inlined:
+				a := src[s+4:]
+				b := src[t+4:]
+				for len(a) >= 8 {
+					if diff := binary.LittleEndian.Uint64(a) ^ binary.LittleEndian.Uint64(b); diff != 0 {
+						l += int32(bits.TrailingZeros64(diff) >> 3)
+						break
+					}
+					l += 8
+					a = a[8:]
+					b = b[8:]
+				}
+				if len(a) < 8 {
+					b = b[:len(a)]
+					for i := range a {
+						if a[i] != b[i] {
+							break
+						}
+						l++
+					}
+				}
+			}
 
 			// Extend backwards
 			for t > 0 && s > nextEmit && src[t-1] == src[s-1] {
@@ -129,7 +158,28 @@ func (e *fastEncL1) Encode(dst *tokens, src []byte) {
 			}
 
 			// Save the match found
-			dst.AddMatchLong(l, uint32(s-t-baseMatchOffset))
+			if false {
+				dst.AddMatchLong(l, uint32(s-t-baseMatchOffset))
+			} else {
+				// Inlined...
+				xoffset := uint32(s - t - baseMatchOffset)
+				xlength := l
+				oc := offsetCode(xoffset)
+				xoffset |= oc << 16
+				for xlength > 0 {
+					xl := xlength
+					if xl > 258 {
+						// We need to have at least baseMatchLength left over for next loop.
+						xl = 258 - baseMatchLength
+					}
+					xlength -= xl
+					xl -= baseMatchLength
+					dst.extraHist[lengthCodes1[uint8(xl)]]++
+					dst.offHist[oc]++
+					dst.tokens[dst.n] = token(matchType | uint32(xl)<<lengthShift | xoffset)
+					dst.n++
+				}
+			}
 			s += l
 			nextEmit = s
 			if nextS >= s {

vendor/github.com/klauspost/compress/flate/level3.go

@@ -5,7 +5,7 @@ import "fmt"
 // fastEncL3
 type fastEncL3 struct {
 	fastGen
-	table [tableSize]tableEntryPrev
+	table [1 << 16]tableEntryPrev
 }
 
 // Encode uses a similar algorithm to level 2, will check up to two candidates.
@@ -13,6 +13,8 @@ func (e *fastEncL3) Encode(dst *tokens, src []byte) {
 	const (
 		inputMargin            = 8 - 1
 		minNonLiteralBlockSize = 1 + 1 + inputMargin
+		tableBits              = 16
+		tableSize              = 1 << tableBits
 	)
 
 	if debugDeflate && e.cur < 0 {
@@ -73,7 +75,7 @@ func (e *fastEncL3) Encode(dst *tokens, src []byte) {
 		nextS := s
 		var candidate tableEntry
 		for {
-			nextHash := hash(cv)
+			nextHash := hash4u(cv, tableBits)
 			s = nextS
 			nextS = s + 1 + (s-nextEmit)>>skipLog
 			if nextS > sLimit {
@@ -156,7 +158,7 @@ func (e *fastEncL3) Encode(dst *tokens, src []byte) {
 				// Index first pair after match end.
 				if int(t+4) < len(src) && t > 0 {
 					cv := load3232(src, t)
-					nextHash := hash(cv)
+					nextHash := hash4u(cv, tableBits)
 					e.table[nextHash] = tableEntryPrev{
 						Prev: e.table[nextHash].Cur,
 						Cur:  tableEntry{offset: e.cur + t},
@@ -165,30 +167,31 @@ func (e *fastEncL3) Encode(dst *tokens, src []byte) {
 				goto emitRemainder
 			}
 
-			// We could immediately start working at s now, but to improve
-			// compression we first update the hash table at s-3 to s.
-			x := load6432(src, s-3)
-			prevHash := hash(uint32(x))
-			e.table[prevHash] = tableEntryPrev{
-				Prev: e.table[prevHash].Cur,
-				Cur:  tableEntry{offset: e.cur + s - 3},
+			// Store every 5th hash in-between.
+			for i := s - l + 2; i < s-5; i += 5 {
+				nextHash := hash4u(load3232(src, i), tableBits)
+				e.table[nextHash] = tableEntryPrev{
+					Prev: e.table[nextHash].Cur,
+					Cur:  tableEntry{offset: e.cur + i}}
 			}
-			x >>= 8
-			prevHash = hash(uint32(x))
+			// We could immediately start working at s now, but to improve
+			// compression we first update the hash table at s-2 to s.
+			x := load6432(src, s-2)
+			prevHash := hash4u(uint32(x), tableBits)
 
 			e.table[prevHash] = tableEntryPrev{
 				Prev: e.table[prevHash].Cur,
 				Cur:  tableEntry{offset: e.cur + s - 2},
 			}
 			x >>= 8
-			prevHash = hash(uint32(x))
+			prevHash = hash4u(uint32(x), tableBits)
 
 			e.table[prevHash] = tableEntryPrev{
 				Prev: e.table[prevHash].Cur,
 				Cur:  tableEntry{offset: e.cur + s - 1},
 			}
 			x >>= 8
-			currHash := hash(uint32(x))
+			currHash := hash4u(uint32(x), tableBits)
 			candidates := e.table[currHash]
 			cv = uint32(x)
 			e.table[currHash] = tableEntryPrev{
@@ -200,15 +203,15 @@ func (e *fastEncL3) Encode(dst *tokens, src []byte) {
 			candidate = candidates.Cur
 			minOffset := e.cur + s - (maxMatchOffset - 4)
 
-			if candidate.offset > minOffset && cv != load3232(src, candidate.offset-e.cur) {
-				// We only check if value mismatches.
-				// Offset will always be invalid in other cases.
+			if candidate.offset > minOffset {
+				if cv == load3232(src, candidate.offset-e.cur) {
+					// Found a match...
+					continue
+				}
 				candidate = candidates.Prev
 				if candidate.offset > minOffset && cv == load3232(src, candidate.offset-e.cur) {
-					offset := s - (candidate.offset - e.cur)
-					if offset <= maxMatchOffset {
-						continue
-					}
+					// Match at prev...
+					continue
 				}
 			}
 			cv = uint32(x >> 8)

vendor/github.com/klauspost/compress/flate/token.go

@@ -13,11 +13,10 @@ import (
 )
 
 const (
-	// From top
-	// 2 bits:   type   0 = literal  1=EOF  2=Match   3=Unused
-	// 8 bits:   xlength = length - MIN_MATCH_LENGTH
-	// 5 bits    offsetcode
-	// 16 bits   xoffset = offset - MIN_OFFSET_SIZE, or literal
+	// bits 0-16  	xoffset = offset - MIN_OFFSET_SIZE, or literal - 16 bits
+	// bits 16-22	offsetcode - 5 bits
+	// bits 22-30   xlength = length - MIN_MATCH_LENGTH - 8 bits
+	// bits 30-32   type   0 = literal  1=EOF  2=Match   3=Unused - 2 bits
 	lengthShift         = 22
 	offsetMask          = 1<<lengthShift - 1
 	typeMask            = 3 << 30
@@ -276,7 +275,7 @@ func (t *tokens) AddMatch(xlength uint32, xoffset uint32) {
 	xoffset |= oCode << 16
 
 	t.extraHist[lengthCodes1[uint8(xlength)]]++
-	t.offHist[oCode]++
+	t.offHist[oCode&31]++
 	t.tokens[t.n] = token(matchType | xlength<<lengthShift | xoffset)
 	t.n++
 }
@@ -300,7 +299,7 @@ func (t *tokens) AddMatchLong(xlength int32, xoffset uint32) {
 		xlength -= xl
 		xl -= baseMatchLength
 		t.extraHist[lengthCodes1[uint8(xl)]]++
-		t.offHist[oc]++
+		t.offHist[oc&31]++
 		t.tokens[t.n] = token(matchType | uint32(xl)<<lengthShift | xoffset)
 		t.n++
 	}
@@ -356,8 +355,8 @@ func (t token) offset() uint32 { return uint32(t) & offsetMask }
 
 func (t token) length() uint8 { return uint8(t >> lengthShift) }
 
-// The code is never more than 8 bits, but is returned as uint32 for convenience.
-func lengthCode(len uint8) uint32 { return uint32(lengthCodes[len]) }
+// Convert length to code.
+func lengthCode(len uint8) uint8 { return lengthCodes[len] }
 
 // Returns the offset code corresponding to a specific offset
 func offsetCode(off uint32) uint32 {

vendor/github.com/klauspost/compress/huff0/decompress.go

@@ -4,6 +4,7 @@ import (
 	"errors"
 	"fmt"
 	"io"
+	"sync"
 
 	"github.com/klauspost/compress/fse"
 )
@@ -216,6 +217,7 @@ func (s *Scratch) Decoder() *Decoder {
 	return &Decoder{
 		dt:             s.dt,
 		actualTableLog: s.actualTableLog,
+		bufs:           &s.decPool,
 	}
 }
 
@@ -223,6 +225,15 @@ func (s *Scratch) Decoder() *Decoder {
 type Decoder struct {
 	dt             dTable
 	actualTableLog uint8
+	bufs           *sync.Pool
+}
+
+func (d *Decoder) buffer() *[4][256]byte {
+	buf, ok := d.bufs.Get().(*[4][256]byte)
+	if ok {
+		return buf
+	}
+	return &[4][256]byte{}
 }
 
 // Decompress1X will decompress a 1X encoded stream.
@@ -249,7 +260,8 @@ func (d *Decoder) Decompress1X(dst, src []byte) ([]byte, error) {
 	dt := d.dt.single[:tlSize]
 
 	// Use temp table to avoid bound checks/append penalty.
-	var buf [256]byte
+	bufs := d.buffer()
+	buf := &bufs[0]
 	var off uint8
 
 	for br.off >= 8 {
@@ -277,6 +289,7 @@ func (d *Decoder) Decompress1X(dst, src []byte) ([]byte, error) {
 		if off == 0 {
 			if len(dst)+256 > maxDecodedSize {
 				br.close()
+				d.bufs.Put(bufs)
 				return nil, ErrMaxDecodedSizeExceeded
 			}
 			dst = append(dst, buf[:]...)
@@ -284,6 +297,7 @@ func (d *Decoder) Decompress1X(dst, src []byte) ([]byte, error) {
 	}
 
 	if len(dst)+int(off) > maxDecodedSize {
+		d.bufs.Put(bufs)
 		br.close()
 		return nil, ErrMaxDecodedSizeExceeded
 	}
@@ -310,6 +324,7 @@ func (d *Decoder) Decompress1X(dst, src []byte) ([]byte, error) {
 			}
 		}
 		if len(dst) >= maxDecodedSize {
+			d.bufs.Put(bufs)
 			br.close()
 			return nil, ErrMaxDecodedSizeExceeded
 		}
@@ -319,6 +334,7 @@ func (d *Decoder) Decompress1X(dst, src []byte) ([]byte, error) {
 		bitsLeft -= nBits
 		dst = append(dst, uint8(v.entry>>8))
 	}
+	d.bufs.Put(bufs)
 	return dst, br.close()
 }
 
@@ -341,7 +357,8 @@ func (d *Decoder) decompress1X8Bit(dst, src []byte) ([]byte, error) {
 	dt := d.dt.single[:256]
 
 	// Use temp table to avoid bound checks/append penalty.
-	var buf [256]byte
+	bufs := d.buffer()
+	buf := &bufs[0]
 	var off uint8
 
 	switch d.actualTableLog {
@@ -369,6 +386,7 @@ func (d *Decoder) decompress1X8Bit(dst, src []byte) ([]byte, error) {
 			if off == 0 {
 				if len(dst)+256 > maxDecodedSize {
 					br.close()
+					d.bufs.Put(bufs)
 					return nil, ErrMaxDecodedSizeExceeded
 				}
 				dst = append(dst, buf[:]...)
@@ -398,6 +416,7 @@ func (d *Decoder) decompress1X8Bit(dst, src []byte) ([]byte, error) {
 			if off == 0 {
 				if len(dst)+256 > maxDecodedSize {
 					br.close()
+					d.bufs.Put(bufs)
 					return nil, ErrMaxDecodedSizeExceeded
 				}
 				dst = append(dst, buf[:]...)
@@ -426,6 +445,7 @@ func (d *Decoder) decompress1X8Bit(dst, src []byte) ([]byte, error) {
 			off += 4
 			if off == 0 {
 				if len(dst)+256 > maxDecodedSize {
+					d.bufs.Put(bufs)
 					br.close()
 					return nil, ErrMaxDecodedSizeExceeded
 				}
@@ -455,6 +475,7 @@ func (d *Decoder) decompress1X8Bit(dst, src []byte) ([]byte, error) {
 			off += 4
 			if off == 0 {
 				if len(dst)+256 > maxDecodedSize {
+					d.bufs.Put(bufs)
 					br.close()
 					return nil, ErrMaxDecodedSizeExceeded
 				}
@@ -484,6 +505,7 @@ func (d *Decoder) decompress1X8Bit(dst, src []byte) ([]byte, error) {
 			off += 4
 			if off == 0 {
 				if len(dst)+256 > maxDecodedSize {
+					d.bufs.Put(bufs)
 					br.close()
 					return nil, ErrMaxDecodedSizeExceeded
 				}
@@ -513,6 +535,7 @@ func (d *Decoder) decompress1X8Bit(dst, src []byte) ([]byte, error) {
 			off += 4
 			if off == 0 {
 				if len(dst)+256 > maxDecodedSize {
+					d.bufs.Put(bufs)
 					br.close()
 					return nil, ErrMaxDecodedSizeExceeded
 				}
@@ -542,6 +565,7 @@ func (d *Decoder) decompress1X8Bit(dst, src []byte) ([]byte, error) {
 			off += 4
 			if off == 0 {
 				if len(dst)+256 > maxDecodedSize {
+					d.bufs.Put(bufs)
 					br.close()
 					return nil, ErrMaxDecodedSizeExceeded
 				}
@@ -571,6 +595,7 @@ func (d *Decoder) decompress1X8Bit(dst, src []byte) ([]byte, error) {
 			off += 4
 			if off == 0 {
 				if len(dst)+256 > maxDecodedSize {
+					d.bufs.Put(bufs)
 					br.close()
 					return nil, ErrMaxDecodedSizeExceeded
 				}
@@ -578,10 +603,12 @@ func (d *Decoder) decompress1X8Bit(dst, src []byte) ([]byte, error) {
 			}
 		}
 	default:
+		d.bufs.Put(bufs)
 		return nil, fmt.Errorf("invalid tablelog: %d", d.actualTableLog)
 	}
 
 	if len(dst)+int(off) > maxDecodedSize {
+		d.bufs.Put(bufs)
 		br.close()
 		return nil, ErrMaxDecodedSizeExceeded
 	}
@@ -601,6 +628,7 @@ func (d *Decoder) decompress1X8Bit(dst, src []byte) ([]byte, error) {
 		}
 		if len(dst) >= maxDecodedSize {
 			br.close()
+			d.bufs.Put(bufs)
 			return nil, ErrMaxDecodedSizeExceeded
 		}
 		v := dt[br.peekByteFast()>>shift]
@@ -609,6 +637,7 @@ func (d *Decoder) decompress1X8Bit(dst, src []byte) ([]byte, error) {
 		bitsLeft -= int8(nBits)
 		dst = append(dst, uint8(v.entry>>8))
 	}
+	d.bufs.Put(bufs)
 	return dst, br.close()
 }
 
@@ -628,7 +657,8 @@ func (d *Decoder) decompress1X8BitExactly(dst, src []byte) ([]byte, error) {
 	dt := d.dt.single[:256]
 
 	// Use temp table to avoid bound checks/append penalty.
-	var buf [256]byte
+	bufs := d.buffer()
+	buf := &bufs[0]
 	var off uint8
 
 	const shift = 56
@@ -655,6 +685,7 @@ func (d *Decoder) decompress1X8BitExactly(dst, src []byte) ([]byte, error) {
 		off += 4
 		if off == 0 {
 			if len(dst)+256 > maxDecodedSize {
+				d.bufs.Put(bufs)
 				br.close()
 				return nil, ErrMaxDecodedSizeExceeded
 			}
@@ -663,6 +694,7 @@ func (d *Decoder) decompress1X8BitExactly(dst, src []byte) ([]byte, error) {
 	}
 
 	if len(dst)+int(off) > maxDecodedSize {
+		d.bufs.Put(bufs)
 		br.close()
 		return nil, ErrMaxDecodedSizeExceeded
 	}
@@ -679,6 +711,7 @@ func (d *Decoder) decompress1X8BitExactly(dst, src []byte) ([]byte, error) {
 			}
 		}
 		if len(dst) >= maxDecodedSize {
+			d.bufs.Put(bufs)
 			br.close()
 			return nil, ErrMaxDecodedSizeExceeded
 		}
@@ -688,6 +721,7 @@ func (d *Decoder) decompress1X8BitExactly(dst, src []byte) ([]byte, error) {
 		bitsLeft -= int8(nBits)
 		dst = append(dst, uint8(v.entry>>8))
 	}
+	d.bufs.Put(bufs)
 	return dst, br.close()
 }
 
@@ -735,12 +769,12 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) {
 	single := d.dt.single[:tlSize]
 
 	// Use temp table to avoid bound checks/append penalty.
-	var buf [256]byte
+	buf := d.buffer()
 	var off uint8
 	var decoded int
 
 	// Decode 2 values from each decoder/loop.
-	const bufoff = 256 / 4
+	const bufoff = 256
 	for {
 		if br[0].off < 4 || br[1].off < 4 || br[2].off < 4 || br[3].off < 4 {
 			break
@@ -758,8 +792,8 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) {
 			v2 := single[val2&tlMask]
 			br[stream].advance(uint8(v.entry))
 			br[stream2].advance(uint8(v2.entry))
-			buf[off+bufoff*stream] = uint8(v.entry >> 8)
-			buf[off+bufoff*stream2] = uint8(v2.entry >> 8)
+			buf[stream][off] = uint8(v.entry >> 8)
+			buf[stream2][off] = uint8(v2.entry >> 8)
 
 			val = br[stream].peekBitsFast(d.actualTableLog)
 			val2 = br[stream2].peekBitsFast(d.actualTableLog)
@@ -767,8 +801,8 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) {
 			v2 = single[val2&tlMask]
 			br[stream].advance(uint8(v.entry))
 			br[stream2].advance(uint8(v2.entry))
-			buf[off+bufoff*stream+1] = uint8(v.entry >> 8)
-			buf[off+bufoff*stream2+1] = uint8(v2.entry >> 8)
+			buf[stream][off+1] = uint8(v.entry >> 8)
+			buf[stream2][off+1] = uint8(v2.entry >> 8)
 		}
 
 		{
@@ -783,8 +817,8 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) {
 			v2 := single[val2&tlMask]
 			br[stream].advance(uint8(v.entry))
 			br[stream2].advance(uint8(v2.entry))
-			buf[off+bufoff*stream] = uint8(v.entry >> 8)
-			buf[off+bufoff*stream2] = uint8(v2.entry >> 8)
+			buf[stream][off] = uint8(v.entry >> 8)
+			buf[stream2][off] = uint8(v2.entry >> 8)
 
 			val = br[stream].peekBitsFast(d.actualTableLog)
 			val2 = br[stream2].peekBitsFast(d.actualTableLog)
@@ -792,25 +826,26 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) {
 			v2 = single[val2&tlMask]
 			br[stream].advance(uint8(v.entry))
 			br[stream2].advance(uint8(v2.entry))
-			buf[off+bufoff*stream+1] = uint8(v.entry >> 8)
-			buf[off+bufoff*stream2+1] = uint8(v2.entry >> 8)
+			buf[stream][off+1] = uint8(v.entry >> 8)
+			buf[stream2][off+1] = uint8(v2.entry >> 8)
 		}
 
 		off += 2
 
-		if off == bufoff {
+		if off == 0 {
 			if bufoff > dstEvery {
+				d.bufs.Put(buf)
 				return nil, errors.New("corruption detected: stream overrun 1")
 			}
-			copy(out, buf[:bufoff])
-			copy(out[dstEvery:], buf[bufoff:bufoff*2])
-			copy(out[dstEvery*2:], buf[bufoff*2:bufoff*3])
-			copy(out[dstEvery*3:], buf[bufoff*3:bufoff*4])
-			off = 0
+			copy(out, buf[0][:])
+			copy(out[dstEvery:], buf[1][:])
+			copy(out[dstEvery*2:], buf[2][:])
+			copy(out[dstEvery*3:], buf[3][:])
 			out = out[bufoff:]
-			decoded += 256
+			decoded += bufoff * 4
 			// There must at least be 3 buffers left.
 			if len(out) < dstEvery*3 {
+				d.bufs.Put(buf)
 				return nil, errors.New("corruption detected: stream overrun 2")
 			}
 		}
@@ -818,12 +853,13 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) {
 	if off > 0 {
 		ioff := int(off)
 		if len(out) < dstEvery*3+ioff {
+			d.bufs.Put(buf)
 			return nil, errors.New("corruption detected: stream overrun 3")
 		}
-		copy(out, buf[:off])
-		copy(out[dstEvery:dstEvery+ioff], buf[bufoff:bufoff*2])
-		copy(out[dstEvery*2:dstEvery*2+ioff], buf[bufoff*2:bufoff*3])
-		copy(out[dstEvery*3:dstEvery*3+ioff], buf[bufoff*3:bufoff*4])
+		copy(out, buf[0][:off])
+		copy(out[dstEvery:], buf[1][:off])
+		copy(out[dstEvery*2:], buf[2][:off])
+		copy(out[dstEvery*3:], buf[3][:off])
 		decoded += int(off) * 4
 		out = out[off:]
 	}
@@ -853,6 +889,7 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) {
 			}
 			// end inline...
 			if offset >= len(out) {
+				d.bufs.Put(buf)
 				return nil, errors.New("corruption detected: stream overrun 4")
 			}
 
@@ -871,6 +908,7 @@ func (d *Decoder) Decompress4X(dst, src []byte) ([]byte, error) {
 			return nil, err
 		}
 	}
+	d.bufs.Put(buf)
 	if dstSize != decoded {
 		return nil, errors.New("corruption detected: short output block")
 	}
@@ -916,12 +954,12 @@ func (d *Decoder) decompress4X8bit(dst, src []byte) ([]byte, error) {
 	single := d.dt.single[:tlSize]
 
 	// Use temp table to avoid bound checks/append penalty.
-	var buf [256]byte
+	buf := d.buffer()
 	var off uint8
 	var decoded int
 
 	// Decode 4 values from each decoder/loop.
-	const bufoff = 256 / 4
+	const bufoff = 256
 	for {
 		if br[0].off < 4 || br[1].off < 4 || br[2].off < 4 || br[3].off < 4 {
 			break
@@ -942,8 +980,8 @@ func (d *Decoder) decompress4X8bit(dst, src []byte) ([]byte, error) {
 			br1.value <<= v & 63
 			br2.bitsRead += uint8(v2)
 			br2.value <<= v2 & 63
-			buf[off+bufoff*stream] = uint8(v >> 8)
-			buf[off+bufoff*stream2] = uint8(v2 >> 8)
+			buf[stream][off] = uint8(v >> 8)
+			buf[stream2][off] = uint8(v2 >> 8)
 
 			v = single[uint8(br1.value>>shift)].entry
 			v2 = single[uint8(br2.value>>shift)].entry
@@ -951,8 +989,8 @@ func (d *Decoder) decompress4X8bit(dst, src []byte) ([]byte, error) {
 			br1.value <<= v & 63
 			br2.bitsRead += uint8(v2)
 			br2.value <<= v2 & 63
-			buf[off+bufoff*stream+1] = uint8(v >> 8)
-			buf[off+bufoff*stream2+1] = uint8(v2 >> 8)
+			buf[stream][off+1] = uint8(v >> 8)
+			buf[stream2][off+1] = uint8(v2 >> 8)
 
 			v = single[uint8(br1.value>>shift)].entry
 			v2 = single[uint8(br2.value>>shift)].entry
@@ -960,8 +998,8 @@ func (d *Decoder) decompress4X8bit(dst, src []byte) ([]byte, error) {
 			br1.value <<= v & 63
 			br2.bitsRead += uint8(v2)
 			br2.value <<= v2 & 63
-			buf[off+bufoff*stream+2] = uint8(v >> 8)
-			buf[off+bufoff*stream2+2] = uint8(v2 >> 8)
+			buf[stream][off+2] = uint8(v >> 8)
+			buf[stream2][off+2] = uint8(v2 >> 8)
 
 			v = single[uint8(br1.value>>shift)].entry
 			v2 = single[uint8(br2.value>>shift)].entry
@@ -969,8 +1007,8 @@ func (d *Decoder) decompress4X8bit(dst, src []byte) ([]byte, error) {
 			br1.value <<= v & 63
 			br2.bitsRead += uint8(v2)
 			br2.value <<= v2 & 63
-			buf[off+bufoff*stream2+3] = uint8(v2 >> 8)
-			buf[off+bufoff*stream+3] = uint8(v >> 8)
+			buf[stream][off+3] = uint8(v >> 8)
+			buf[stream2][off+3] = uint8(v2 >> 8)
 		}
 
 		{
@@ -987,8 +1025,8 @@ func (d *Decoder) decompress4X8bit(dst, src []byte) ([]byte, error) {
 			br1.value <<= v & 63
 			br2.bitsRead += uint8(v2)
 			br2.value <<= v2 & 63
-			buf[off+bufoff*stream] = uint8(v >> 8)
-			buf[off+bufoff*stream2] = uint8(v2 >> 8)
+			buf[stream][off] = uint8(v >> 8)
+			buf[stream2][off] = uint8(v2 >> 8)
 
 			v = single[uint8(br1.value>>shift)].entry
 			v2 = single[uint8(br2.value>>shift)].entry
@@ -996,8 +1034,8 @@ func (d *Decoder) decompress4X8bit(dst, src []byte) ([]byte, error) {
 			br1.value <<= v & 63
 			br2.bitsRead += uint8(v2)
 			br2.value <<= v2 & 63
-			buf[off+bufoff*stream+1] = uint8(v >> 8)
-			buf[off+bufoff*stream2+1] = uint8(v2 >> 8)
+			buf[stream][off+1] = uint8(v >> 8)
+			buf[stream2][off+1] = uint8(v2 >> 8)
 
 			v = single[uint8(br1.value>>shift)].entry
 			v2 = single[uint8(br2.value>>shift)].entry
@@ -1005,8 +1043,8 @@ func (d *Decoder) decompress4X8bit(dst, src []byte) ([]byte, error) {
 			br1.value <<= v & 63
 			br2.bitsRead += uint8(v2)
 			br2.value <<= v2 & 63
-			buf[off+bufoff*stream+2] = uint8(v >> 8)
-			buf[off+bufoff*stream2+2] = uint8(v2 >> 8)
+			buf[stream][off+2] = uint8(v >> 8)
+			buf[stream2][off+2] = uint8(v2 >> 8)
 
 			v = single[uint8(br1.value>>shift)].entry
 			v2 = single[uint8(br2.value>>shift)].entry
@@ -1014,25 +1052,26 @@ func (d *Decoder) decompress4X8bit(dst, src []byte) ([]byte, error) {
 			br1.value <<= v & 63
 			br2.bitsRead += uint8(v2)
 			br2.value <<= v2 & 63
-			buf[off+bufoff*stream2+3] = uint8(v2 >> 8)
-			buf[off+bufoff*stream+3] = uint8(v >> 8)
+			buf[stream][off+3] = uint8(v >> 8)
+			buf[stream2][off+3] = uint8(v2 >> 8)
 		}
 
 		off += 4
 
-		if off == bufoff {
+		if off == 0 {
 			if bufoff > dstEvery {
+				d.bufs.Put(buf)
 				return nil, errors.New("corruption detected: stream overrun 1")
 			}
-			copy(out, buf[:bufoff])
-			copy(out[dstEvery:], buf[bufoff:bufoff*2])
-			copy(out[dstEvery*2:], buf[bufoff*2:bufoff*3])
-			copy(out[dstEvery*3:], buf[bufoff*3:bufoff*4])
-			off = 0
+			copy(out, buf[0][:])
+			copy(out[dstEvery:], buf[1][:])
+			copy(out[dstEvery*2:], buf[2][:])
+			copy(out[dstEvery*3:], buf[3][:])
 			out = out[bufoff:]
-			decoded += 256
+			decoded += bufoff * 4
 			// There must at least be 3 buffers left.
 			if len(out) < dstEvery*3 {
+				d.bufs.Put(buf)
 				return nil, errors.New("corruption detected: stream overrun 2")
 			}
 		}
@@ -1040,12 +1079,13 @@ func (d *Decoder) decompress4X8bit(dst, src []byte) ([]byte, error) {
 	if off > 0 {
 		ioff := int(off)
 		if len(out) < dstEvery*3+ioff {
+			d.bufs.Put(buf)
 			return nil, errors.New("corruption detected: stream overrun 3")
 		}
-		copy(out, buf[:off])
-		copy(out[dstEvery:dstEvery+ioff], buf[bufoff:bufoff*2])
-		copy(out[dstEvery*2:dstEvery*2+ioff], buf[bufoff*2:bufoff*3])
-		copy(out[dstEvery*3:dstEvery*3+ioff], buf[bufoff*3:bufoff*4])
+		copy(out, buf[0][:off])
+		copy(out[dstEvery:], buf[1][:off])
+		copy(out[dstEvery*2:], buf[2][:off])
+		copy(out[dstEvery*3:], buf[3][:off])
 		decoded += int(off) * 4
 		out = out[off:]
 	}
@@ -1057,6 +1097,7 @@ func (d *Decoder) decompress4X8bit(dst, src []byte) ([]byte, error) {
 		bitsLeft := int(br.off*8) + int(64-br.bitsRead)
 		for bitsLeft > 0 {
 			if br.finished() {
+				d.bufs.Put(buf)
 				return nil, io.ErrUnexpectedEOF
 			}
 			if br.bitsRead >= 56 {
@@ -1077,6 +1118,7 @@ func (d *Decoder) decompress4X8bit(dst, src []byte) ([]byte, error) {
 			}
 			// end inline...
 			if offset >= len(out) {
+				d.bufs.Put(buf)
 				return nil, errors.New("corruption detected: stream overrun 4")
 			}
 
@@ -1091,9 +1133,11 @@ func (d *Decoder) decompress4X8bit(dst, src []byte) ([]byte, error) {
 		decoded += offset - dstEvery*i
 		err = br.close()
 		if err != nil {
+			d.bufs.Put(buf)
 			return nil, err
 		}
 	}
+	d.bufs.Put(buf)
 	if dstSize != decoded {
 		return nil, errors.New("corruption detected: short output block")
 	}
@@ -1135,12 +1179,12 @@ func (d *Decoder) decompress4X8bitExactly(dst, src []byte) ([]byte, error) {
 	single := d.dt.single[:tlSize]
 
 	// Use temp table to avoid bound checks/append penalty.
-	var buf [256]byte
+	buf := d.buffer()
 	var off uint8
 	var decoded int
 
 	// Decode 4 values from each decoder/loop.
-	const bufoff = 256 / 4
+	const bufoff = 256
 	for {
 		if br[0].off < 4 || br[1].off < 4 || br[2].off < 4 || br[3].off < 4 {
 			break
@@ -1150,104 +1194,109 @@ func (d *Decoder) decompress4X8bitExactly(dst, src []byte) ([]byte, error) {
 			// Interleave 2 decodes.
 			const stream = 0
 			const stream2 = 1
-			br[stream].fillFast()
-			br[stream2].fillFast()
+			br1 := &br[stream]
+			br2 := &br[stream2]
+			br1.fillFast()
+			br2.fillFast()
 
-			v := single[uint8(br[stream].value>>shift)].entry
-			v2 := single[uint8(br[stream2].value>>shift)].entry
-			br[stream].bitsRead += uint8(v)
-			br[stream].value <<= v & 63
-			br[stream2].bitsRead += uint8(v2)
-			br[stream2].value <<= v2 & 63
-			buf[off+bufoff*stream] = uint8(v >> 8)
-			buf[off+bufoff*stream2] = uint8(v2 >> 8)
+			v := single[uint8(br1.value>>shift)].entry
+			v2 := single[uint8(br2.value>>shift)].entry
+			br1.bitsRead += uint8(v)
+			br1.value <<= v & 63
+			br2.bitsRead += uint8(v2)
+			br2.value <<= v2 & 63
+			buf[stream][off] = uint8(v >> 8)
+			buf[stream2][off] = uint8(v2 >> 8)
 
-			v = single[uint8(br[stream].value>>shift)].entry
-			v2 = single[uint8(br[stream2].value>>shift)].entry
-			br[stream].bitsRead += uint8(v)
-			br[stream].value <<= v & 63
-			br[stream2].bitsRead += uint8(v2)
-			br[stream2].value <<= v2 & 63
-			buf[off+bufoff*stream+1] = uint8(v >> 8)
-			buf[off+bufoff*stream2+1] = uint8(v2 >> 8)
+			v = single[uint8(br1.value>>shift)].entry
+			v2 = single[uint8(br2.value>>shift)].entry
+			br1.bitsRead += uint8(v)
+			br1.value <<= v & 63
+			br2.bitsRead += uint8(v2)
+			br2.value <<= v2 & 63
+			buf[stream][off+1] = uint8(v >> 8)
+			buf[stream2][off+1] = uint8(v2 >> 8)
 
-			v = single[uint8(br[stream].value>>shift)].entry
-			v2 = single[uint8(br[stream2].value>>shift)].entry
-			br[stream].bitsRead += uint8(v)
-			br[stream].value <<= v & 63
-			br[stream2].bitsRead += uint8(v2)
-			br[stream2].value <<= v2 & 63
-			buf[off+bufoff*stream+2] = uint8(v >> 8)
-			buf[off+bufoff*stream2+2] = uint8(v2 >> 8)
+			v = single[uint8(br1.value>>shift)].entry
+			v2 = single[uint8(br2.value>>shift)].entry
+			br1.bitsRead += uint8(v)
+			br1.value <<= v & 63
+			br2.bitsRead += uint8(v2)
+			br2.value <<= v2 & 63
+			buf[stream][off+2] = uint8(v >> 8)
+			buf[stream2][off+2] = uint8(v2 >> 8)
 
-			v = single[uint8(br[stream].value>>shift)].entry
-			v2 = single[uint8(br[stream2].value>>shift)].entry
-			br[stream].bitsRead += uint8(v)
-			br[stream].value <<= v & 63
-			br[stream2].bitsRead += uint8(v2)
-			br[stream2].value <<= v2 & 63
-			buf[off+bufoff*stream+3] = uint8(v >> 8)
-			buf[off+bufoff*stream2+3] = uint8(v2 >> 8)
+			v = single[uint8(br1.value>>shift)].entry
+			v2 = single[uint8(br2.value>>shift)].entry
+			br1.bitsRead += uint8(v)
+			br1.value <<= v & 63
+			br2.bitsRead += uint8(v2)
+			br2.value <<= v2 & 63
+			buf[stream][off+3] = uint8(v >> 8)
+			buf[stream2][off+3] = uint8(v2 >> 8)
 		}
 
 		{
 			const stream = 2
 			const stream2 = 3
-			br[stream].fillFast()
-			br[stream2].fillFast()
+			br1 := &br[stream]
+			br2 := &br[stream2]
+			br1.fillFast()
+			br2.fillFast()
 
-			v := single[uint8(br[stream].value>>shift)].entry
-			v2 := single[uint8(br[stream2].value>>shift)].entry
-			br[stream].bitsRead += uint8(v)
-			br[stream].value <<= v & 63
-			br[stream2].bitsRead += uint8(v2)
-			br[stream2].value <<= v2 & 63
-			buf[off+bufoff*stream] = uint8(v >> 8)
-			buf[off+bufoff*stream2] = uint8(v2 >> 8)
+			v := single[uint8(br1.value>>shift)].entry
+			v2 := single[uint8(br2.value>>shift)].entry
+			br1.bitsRead += uint8(v)
+			br1.value <<= v & 63
+			br2.bitsRead += uint8(v2)
+			br2.value <<= v2 & 63
+			buf[stream][off] = uint8(v >> 8)
+			buf[stream2][off] = uint8(v2 >> 8)
 
-			v = single[uint8(br[stream].value>>shift)].entry
-			v2 = single[uint8(br[stream2].value>>shift)].entry
-			br[stream].bitsRead += uint8(v)
-			br[stream].value <<= v & 63
-			br[stream2].bitsRead += uint8(v2)
-			br[stream2].value <<= v2 & 63
-			buf[off+bufoff*stream+1] = uint8(v >> 8)
-			buf[off+bufoff*stream2+1] = uint8(v2 >> 8)
+			v = single[uint8(br1.value>>shift)].entry
+			v2 = single[uint8(br2.value>>shift)].entry
+			br1.bitsRead += uint8(v)
+			br1.value <<= v & 63
+			br2.bitsRead += uint8(v2)
+			br2.value <<= v2 & 63
+			buf[stream][off+1] = uint8(v >> 8)
+			buf[stream2][off+1] = uint8(v2 >> 8)
 
-			v = single[uint8(br[stream].value>>shift)].entry
-			v2 = single[uint8(br[stream2].value>>shift)].entry
-			br[stream].bitsRead += uint8(v)
-			br[stream].value <<= v & 63
-			br[stream2].bitsRead += uint8(v2)
-			br[stream2].value <<= v2 & 63
-			buf[off+bufoff*stream+2] = uint8(v >> 8)
-			buf[off+bufoff*stream2+2] = uint8(v2 >> 8)
+			v = single[uint8(br1.value>>shift)].entry
+			v2 = single[uint8(br2.value>>shift)].entry
+			br1.bitsRead += uint8(v)
+			br1.value <<= v & 63
+			br2.bitsRead += uint8(v2)
+			br2.value <<= v2 & 63
+			buf[stream][off+2] = uint8(v >> 8)
+			buf[stream2][off+2] = uint8(v2 >> 8)
 
-			v = single[uint8(br[stream].value>>shift)].entry
-			v2 = single[uint8(br[stream2].value>>shift)].entry
-			br[stream].bitsRead += uint8(v)
-			br[stream].value <<= v & 63
-			br[stream2].bitsRead += uint8(v2)
-			br[stream2].value <<= v2 & 63
-			buf[off+bufoff*stream+3] = uint8(v >> 8)
-			buf[off+bufoff*stream2+3] = uint8(v2 >> 8)
+			v = single[uint8(br1.value>>shift)].entry
+			v2 = single[uint8(br2.value>>shift)].entry
+			br1.bitsRead += uint8(v)
+			br1.value <<= v & 63
+			br2.bitsRead += uint8(v2)
+			br2.value <<= v2 & 63
+			buf[stream][off+3] = uint8(v >> 8)
+			buf[stream2][off+3] = uint8(v2 >> 8)
 		}
 
 		off += 4
 
-		if off == bufoff {
+		if off == 0 {
 			if bufoff > dstEvery {
+				d.bufs.Put(buf)
 				return nil, errors.New("corruption detected: stream overrun 1")
 			}
-			copy(out, buf[:bufoff])
-			copy(out[dstEvery:], buf[bufoff:bufoff*2])
-			copy(out[dstEvery*2:], buf[bufoff*2:bufoff*3])
-			copy(out[dstEvery*3:], buf[bufoff*3:bufoff*4])
-			off = 0
+			copy(out, buf[0][:])
+			copy(out[dstEvery:], buf[1][:])
+			copy(out[dstEvery*2:], buf[2][:])
+			copy(out[dstEvery*3:], buf[3][:])
 			out = out[bufoff:]
-			decoded += 256
+			decoded += bufoff * 4
 			// There must at least be 3 buffers left.
 			if len(out) < dstEvery*3 {
+				d.bufs.Put(buf)
 				return nil, errors.New("corruption detected: stream overrun 2")
 			}
 		}
@@ -1257,10 +1306,10 @@ func (d *Decoder) decompress4X8bitExactly(dst, src []byte) ([]byte, error) {
 		if len(out) < dstEvery*3+ioff {
 			return nil, errors.New("corruption detected: stream overrun 3")
 		}
-		copy(out, buf[:off])
-		copy(out[dstEvery:dstEvery+ioff], buf[bufoff:bufoff*2])
-		copy(out[dstEvery*2:dstEvery*2+ioff], buf[bufoff*2:bufoff*3])
-		copy(out[dstEvery*3:dstEvery*3+ioff], buf[bufoff*3:bufoff*4])
+		copy(out, buf[0][:off])
+		copy(out[dstEvery:], buf[1][:off])
+		copy(out[dstEvery*2:], buf[2][:off])
+		copy(out[dstEvery*3:], buf[3][:off])
 		decoded += int(off) * 4
 		out = out[off:]
 	}
@@ -1272,6 +1321,7 @@ func (d *Decoder) decompress4X8bitExactly(dst, src []byte) ([]byte, error) {
 		bitsLeft := int(br.off*8) + int(64-br.bitsRead)
 		for bitsLeft > 0 {
 			if br.finished() {
+				d.bufs.Put(buf)
 				return nil, io.ErrUnexpectedEOF
 			}
 			if br.bitsRead >= 56 {
@@ -1292,6 +1342,7 @@ func (d *Decoder) decompress4X8bitExactly(dst, src []byte) ([]byte, error) {
 			}
 			// end inline...
 			if offset >= len(out) {
+				d.bufs.Put(buf)
 				return nil, errors.New("corruption detected: stream overrun 4")
 			}
 
@@ -1306,9 +1357,11 @@ func (d *Decoder) decompress4X8bitExactly(dst, src []byte) ([]byte, error) {
 		decoded += offset - dstEvery*i
 		err = br.close()
 		if err != nil {
+			d.bufs.Put(buf)
 			return nil, err
 		}
 	}
+	d.bufs.Put(buf)
 	if dstSize != decoded {
 		return nil, errors.New("corruption detected: short output block")
 	}

vendor/github.com/klauspost/compress/huff0/huff0.go

@@ -8,6 +8,7 @@ import (
 	"fmt"
 	"math"
 	"math/bits"
+	"sync"
 
 	"github.com/klauspost/compress/fse"
 )
@@ -116,6 +117,7 @@ type Scratch struct {
 	nodes          []nodeElt
 	tmpOut         [4][]byte
 	fse            *fse.Scratch
+	decPool        sync.Pool // *[4][256]byte buffers.
 	huffWeight     [maxSymbolValue + 1]byte
 }
 

vendor/github.com/klauspost/compress/s2/encodeblock_amd64.go

@@ -1,7 +1,7 @@
 // Code generated by command: go run gen.go -out ../encodeblock_amd64.s -stubs ../encodeblock_amd64.go -pkg=s2. DO NOT EDIT.
 
-//go:build !appengine && !noasm && gc
-// +build !appengine,!noasm,gc
+//go:build !appengine && !noasm && gc && !noasm
+// +build !appengine,!noasm,gc,!noasm
 
 package s2
 

vendor/github.com/klauspost/compress/zip/reader.go

@@ -128,7 +128,6 @@ func (z *Reader) init(r io.ReaderAt, size int64) error {
 		if err != nil {
 			return err
 		}
-		f.readDataDescriptor()
 		z.File = append(z.File, f)
 	}
 	if uint16(len(z.File)) != uint16(end.directoryRecords) { // only compare 16 bits here
@@ -189,10 +188,15 @@ func (f *File) Open() (io.ReadCloser, error) {
 		return nil, ErrAlgorithm
 	}
 	var rc io.ReadCloser = dcomp(r)
+	var desr io.Reader
+	if f.hasDataDescriptor() {
+		desr = io.NewSectionReader(f.zipr, f.headerOffset+bodyOffset+size, dataDescriptorLen)
+	}
 	rc = &checksumReader{
 		rc:   rc,
 		hash: crc32.NewIEEE(),
 		f:    f,
+		desr: desr,
 	}
 	return rc, nil
 }
@@ -208,49 +212,13 @@ func (f *File) OpenRaw() (io.Reader, error) {
 	return r, nil
 }
 
-func (f *File) readDataDescriptor() {
-	if !f.hasDataDescriptor() {
-		return
-	}
-
-	bodyOffset, err := f.findBodyOffset()
-	if err != nil {
-		f.descErr = err
-		return
-	}
-
-	// In section 4.3.9.2 of the spec: "However ZIP64 format MAY be used
-	// regardless of the size of a file.  When extracting, if the zip64
-	// extended information extra field is present for the file the
-	// compressed and uncompressed sizes will be 8 byte values."
-	//
-	// Historically, this package has used the compressed and uncompressed
-	// sizes from the central directory to determine if the package is
-	// zip64.
-	//
-	// For this case we allow either the extra field or sizes to determine
-	// the data descriptor length.
-	zip64 := f.zip64 || f.isZip64()
-	n := int64(dataDescriptorLen)
-	if zip64 {
-		n = dataDescriptor64Len
-	}
-	size := int64(f.CompressedSize64)
-	r := io.NewSectionReader(f.zipr, f.headerOffset+bodyOffset+size, n)
-	dd, err := readDataDescriptor(r, zip64)
-	if err != nil {
-		f.descErr = err
-		return
-	}
-	f.CRC32 = dd.crc32
-}
-
 type checksumReader struct {
 	rc    io.ReadCloser
 	hash  hash.Hash32
 	nread uint64 // number of bytes read so far
 	f     *File
-	err   error // sticky error
+	desr  io.Reader // if non-nil, where to read the data descriptor
+	err   error     // sticky error
 }
 
 func (r *checksumReader) Stat() (fs.FileInfo, error) {
@@ -271,12 +239,12 @@ func (r *checksumReader) Read(b []byte) (n int, err error) {
 		if r.nread != r.f.UncompressedSize64 {
 			return 0, io.ErrUnexpectedEOF
 		}
-		if r.f.hasDataDescriptor() {
-			if r.f.descErr != nil {
-				if r.f.descErr == io.EOF {
+		if r.desr != nil {
+			if err1 := readDataDescriptor(r.desr, r.f); err1 != nil {
+				if err1 == io.EOF {
 					err = io.ErrUnexpectedEOF
 				} else {
-					err = r.f.descErr
+					err = err1
 				}
 			} else if r.hash.Sum32() != r.f.CRC32 {
 				err = ErrChecksum
@@ -488,10 +456,8 @@ parseExtras:
 	return nil
 }
 
-func readDataDescriptor(r io.Reader, zip64 bool) (*dataDescriptor, error) {
-	// Create enough space for the largest possible size
-	var buf [dataDescriptor64Len]byte
-
+func readDataDescriptor(r io.Reader, f *File) error {
+	var buf [dataDescriptorLen]byte
 	// The spec says: "Although not originally assigned a
 	// signature, the value 0x08074b50 has commonly been adopted
 	// as a signature value for the data descriptor record.
@@ -500,9 +466,10 @@ func readDataDescriptor(r io.Reader, zip64 bool) (*dataDescriptor, error) {
 	// descriptors and should account for either case when reading
 	// ZIP files to ensure compatibility."
 	//
-	// First read just those 4 bytes to see if the signature exists.
+	// dataDescriptorLen includes the size of the signature but
+	// first read just those 4 bytes to see if it exists.
 	if _, err := io.ReadFull(r, buf[:4]); err != nil {
-		return nil, err
+		return err
 	}
 	off := 0
 	maybeSig := readBuf(buf[:4])
@@ -511,28 +478,21 @@ func readDataDescriptor(r io.Reader, zip64 bool) (*dataDescriptor, error) {
 		// bytes.
 		off += 4
 	}
-
-	end := dataDescriptorLen - 4
-	if zip64 {
-		end = dataDescriptor64Len - 4
+	if _, err := io.ReadFull(r, buf[off:12]); err != nil {
+		return err
 	}
-	if _, err := io.ReadFull(r, buf[off:end]); err != nil {
-		return nil, err
-	}
-	b := readBuf(buf[:end])
-
-	out := &dataDescriptor{
-		crc32: b.uint32(),
+	b := readBuf(buf[:12])
+	if b.uint32() != f.CRC32 {
+		return ErrChecksum
 	}
 
-	if zip64 {
-		out.compressedSize = b.uint64()
-		out.uncompressedSize = b.uint64()
-	} else {
-		out.compressedSize = uint64(b.uint32())
-		out.uncompressedSize = uint64(b.uint32())
-	}
-	return out, nil
+	// The two sizes that follow here can be either 32 bits or 64 bits
+	// but the spec is not very clear on this and different
+	// interpretations has been made causing incompatibilities. We
+	// already have the sizes from the central directory so we can
+	// just ignore these.
+
+	return nil
 }
 
 func readDirectoryEnd(r io.ReaderAt, size int64) (dir *directoryEnd, err error) {

vendor/github.com/klauspost/compress/zip/struct.go

@@ -393,11 +393,3 @@ func unixModeToFileMode(m uint32) fs.FileMode {
 	}
 	return mode
 }
-
-// dataDescriptor holds the data descriptor that optionally follows the file
-// contents in the zip file.
-type dataDescriptor struct {
-	crc32            uint32
-	compressedSize   uint64
-	uncompressedSize uint64
-}

vendor/github.com/klauspost/compress/zip/writer.go

@@ -438,7 +438,7 @@ func min64(x, y uint64) uint64 {
 
 // Deprecated: CreateHeaderRaw is replaced by CreateRaw (stdlib name).
 func (w *Writer) CreateHeaderRaw(fh *FileHeader) (io.Writer, error) {
-	return w.CreateHeader(fh)
+	return w.CreateRaw(fh)
 }
 
 // CreateRaw adds a file to the zip archive using the provided FileHeader and

vendor/github.com/klauspost/compress/zstd/enc_fast.go

@@ -85,7 +85,7 @@ func (e *fastEncoder) Encode(blk *blockEnc, src []byte) {
 	// TEMPLATE
 	const hashLog = tableBits
 	// seems global, but would be nice to tweak.
-	const kSearchStrength = 7
+	const kSearchStrength = 6
 
 	// nextEmit is where in src the next emitLiteral should start from.
 	nextEmit := s
@@ -334,7 +334,7 @@ func (e *fastEncoder) EncodeNoHist(blk *blockEnc, src []byte) {
 	// TEMPLATE
 	const hashLog = tableBits
 	// seems global, but would be nice to tweak.
-	const kSearchStrength = 8
+	const kSearchStrength = 6
 
 	// nextEmit is where in src the next emitLiteral should start from.
 	nextEmit := s

vendor/github.com/minio/minio-go/v7/README.md

@@ -214,14 +214,10 @@ The full API Reference is available here.
 ### Full Examples : File Object Operations
 * [fputobject.go](https://github.com/minio/minio-go/blob/master/examples/s3/fputobject.go)
 * [fgetobject.go](https://github.com/minio/minio-go/blob/master/examples/s3/fgetobject.go)
-* [fputobject-context.go](https://github.com/minio/minio-go/blob/master/examples/s3/fputobject-context.go)
-* [fgetobject-context.go](https://github.com/minio/minio-go/blob/master/examples/s3/fgetobject-context.go)
 
 ### Full Examples : Object Operations
 * [putobject.go](https://github.com/minio/minio-go/blob/master/examples/s3/putobject.go)
 * [getobject.go](https://github.com/minio/minio-go/blob/master/examples/s3/getobject.go)
-* [putobject-context.go](https://github.com/minio/minio-go/blob/master/examples/s3/putobject-context.go)
-* [getobject-context.go](https://github.com/minio/minio-go/blob/master/examples/s3/getobject-context.go)
 * [statobject.go](https://github.com/minio/minio-go/blob/master/examples/s3/statobject.go)
 * [copyobject.go](https://github.com/minio/minio-go/blob/master/examples/s3/copyobject.go)
 * [removeobject.go](https://github.com/minio/minio-go/blob/master/examples/s3/removeobject.go)

vendor/github.com/minio/minio-go/v7/api-bucket-policy.go

@@ -87,6 +87,11 @@ func (c *Client) removeBucketPolicy(ctx context.Context, bucketName string) erro
 	if err != nil {
 		return err
 	}
+
+	if resp.StatusCode != http.StatusNoContent {
+		return httpRespToErrorResponse(resp, bucketName, "")
+	}
+
 	return nil
 }
 

vendor/github.com/minio/minio-go/v7/api-bucket-replication.go

@@ -246,12 +246,42 @@ func (c *Client) resetBucketReplicationOnTarget(ctx context.Context, bucketName
 	if resp.StatusCode != http.StatusOK {
 		return rinfo, httpRespToErrorResponse(resp, bucketName, "")
 	}
-	respBytes, err := ioutil.ReadAll(resp.Body)
-	if err != nil {
-		return rinfo, err
-	}
 
-	if err := json.Unmarshal(respBytes, &rinfo); err != nil {
+	if err = json.NewDecoder(resp.Body).Decode(&rinfo); err != nil {
+		return rinfo, err
+	}
+	return rinfo, nil
+}
+
+// GetBucketReplicationResyncStatus gets the status of replication resync
+func (c *Client) GetBucketReplicationResyncStatus(ctx context.Context, bucketName, arn string) (rinfo replication.ResyncTargetsInfo, err error) {
+	// Input validation.
+	if err := s3utils.CheckValidBucketName(bucketName); err != nil {
+		return rinfo, err
+	}
+	// Get resources properly escaped and lined up before
+	// using them in http request.
+	urlValues := make(url.Values)
+	urlValues.Set("replication-reset-status", "")
+	if arn != "" {
+		urlValues.Set("arn", arn)
+	}
+	// Execute GET on bucket to get replication config.
+	resp, err := c.executeMethod(ctx, http.MethodGet, requestMetadata{
+		bucketName:  bucketName,
+		queryValues: urlValues,
+	})
+
+	defer closeResponse(resp)
+	if err != nil {
+		return rinfo, err
+	}
+
+	if resp.StatusCode != http.StatusOK {
+		return rinfo, httpRespToErrorResponse(resp, bucketName, "")
+	}
+
+	if err = json.NewDecoder(resp.Body).Decode(&rinfo); err != nil {
 		return rinfo, err
 	}
 	return rinfo, nil

vendor/github.com/minio/minio-go/v7/api-restore.go

@@ -175,7 +175,7 @@ func (c *Client) RestoreObject(ctx context.Context, bucketName, objectName, vers
 	if err != nil {
 		return err
 	}
-	if resp.StatusCode != http.StatusAccepted {
+	if resp.StatusCode != http.StatusAccepted && resp.StatusCode != http.StatusOK {
 		return httpRespToErrorResponse(resp, bucketName, "")
 	}
 	return nil

vendor/github.com/minio/minio-go/v7/api.go

@@ -111,7 +111,7 @@ type Options struct {
 // Global constants.
 const (
 	libraryName    = "minio-go"
-	libraryVersion = "v7.0.21"
+	libraryVersion = "v7.0.23"
 )
 
 // User Agent should always following the below style.
@@ -875,10 +875,14 @@ func (c *Client) makeTargetURL(bucketName, objectName, bucketLocation string, is
 	if h, p, err := net.SplitHostPort(host); err == nil {
 		if scheme == "http" && p == "80" || scheme == "https" && p == "443" {
 			host = h
+			if ip := net.ParseIP(h); ip != nil && ip.To16() != nil {
+				host = "[" + h + "]"
+			}
 		}
 	}
 
 	urlStr := scheme + "://" + host + "/"
+
 	// Make URL only if bucketName is available, otherwise use the
 	// endpoint URL.
 	if bucketName != "" {

vendor/github.com/minio/minio-go/v7/functional_tests.go

@@ -8143,6 +8143,8 @@ func testDecryptedCopyObject() {
 		return
 	}
 
+	defer cleanupBucket(bucketName, c)
+
 	encryption := encrypt.DefaultPBKDF([]byte("correct horse battery staple"), []byte(bucketName+objectName))
 	_, err = c.PutObject(context.Background(), bucketName, objectName, bytes.NewReader(bytes.Repeat([]byte("a"), 1024*1024)), 1024*1024, minio.PutObjectOptions{
 		ServerSideEncryption: encryption,

vendor/github.com/minio/minio-go/v7/pkg/replication/replication.go

@@ -22,6 +22,7 @@ import (
 	"fmt"
 	"strconv"
 	"strings"
+	"time"
 	"unicode/utf8"
 
 	"github.com/rs/xid"
@@ -726,6 +727,21 @@ type ResyncTargetsInfo struct {
 
 // ResyncTarget provides the replica resources and resetID to initiate resync replication.
 type ResyncTarget struct {
-	Arn     string `json:"arn"`
-	ResetID string `json:"resetid"`
+	Arn       string    `json:"arn"`
+	ResetID   string    `json:"resetid"`
+	StartTime time.Time `json:"startTime,omitempty"`
+	EndTime   time.Time `json:"endTime,omitempty"`
+	// Status of resync operation
+	ResyncStatus string `json:"resyncStatus,omitempty"`
+	// Completed size in bytes
+	ReplicatedSize int64 `json:"completedReplicationSize,omitempty"`
+	// Failed size in bytes
+	FailedSize int64 `json:"failedReplicationSize,omitempty"`
+	// Total number of failed operations
+	FailedCount int64 `json:"failedReplicationCount,omitempty"`
+	// Total number of failed operations
+	ReplicatedCount int64 `json:"replicationCount,omitempty"`
+	// Last bucket/object replicated.
+	Bucket string `json:"bucket,omitempty"`
+	Object string `json:"object,omitempty"`
 }

vendor/github.com/minio/minio-go/v7/utils.go

@@ -517,11 +517,11 @@ var md5Pool = sync.Pool{New: func() interface{} { return md5.New() }}
 var sha256Pool = sync.Pool{New: func() interface{} { return sha256.New() }}
 
 func newMd5Hasher() md5simd.Hasher {
-	return hashWrapper{Hash: md5Pool.New().(hash.Hash), isMD5: true}
+	return hashWrapper{Hash: md5Pool.Get().(hash.Hash), isMD5: true}
 }
 
 func newSHA256Hasher() md5simd.Hasher {
-	return hashWrapper{Hash: sha256Pool.New().(hash.Hash), isSHA256: true}
+	return hashWrapper{Hash: sha256Pool.Get().(hash.Hash), isSHA256: true}
 }
 
 // hashWrapper implements the md5simd.Hasher interface.

vendor/github.com/yuin/goldmark/README.md

@@ -387,28 +387,29 @@ blackfriday v2 seems to be the fastest, but as it is not CommonMark compliant, i
 goldmark, meanwhile, builds a clean, extensible AST structure, achieves full compliance with
 CommonMark, and consumes less memory, all while being reasonably fast.
 
+- MBP 2019 13″(i5, 16GB), Go1.17
+
 ```
-goos: darwin
-goarch: amd64
-BenchmarkMarkdown/Blackfriday-v2-12                  326           3465240 ns/op         3298861 B/op      20047 allocs/op
-BenchmarkMarkdown/GoldMark-12                        303           3927494 ns/op         2574809 B/op      13853 allocs/op
-BenchmarkMarkdown/CommonMark-12                      244           4900853 ns/op         2753851 B/op      20527 allocs/op
-BenchmarkMarkdown/Lute-12                            130           9195245 ns/op         9175030 B/op     123534 allocs/op
-BenchmarkMarkdown/GoMarkdown-12                        9         113541994 ns/op         2187472 B/op      22173 allocs/op
+BenchmarkMarkdown/Blackfriday-v2-8                   302           3743747 ns/op         3290445 B/op      20050 allocs/op
+BenchmarkMarkdown/GoldMark-8                         280           4200974 ns/op         2559738 B/op      13435 allocs/op
+BenchmarkMarkdown/CommonMark-8                       226           5283686 ns/op         2702490 B/op      20792 allocs/op
+BenchmarkMarkdown/Lute-8                              12          92652857 ns/op        10602649 B/op      40555 allocs/op
+BenchmarkMarkdown/GoMarkdown-8                        13          81380167 ns/op         2245002 B/op      22889 allocs/op
 ```
 
 ### against cmark (CommonMark reference implementation written in C)
 
+- MBP 2019 13″(i5, 16GB), Go1.17
+
 ```
 ----------- cmark -----------
 file: _data.md
 iteration: 50
-average: 0.0037760639 sec
-go run ./goldmark_benchmark.go
+average: 0.0044073057 sec
 ------- goldmark -------
 file: _data.md
 iteration: 50
-average: 0.0040964230 sec
+average: 0.0041611990 sec
 ```
 
 As you can see, goldmark's performance is on par with cmark's.

vendor/github.com/yuin/goldmark/ast/block.go

@@ -87,7 +87,20 @@ func (n *Document) Meta() map[string]interface{} {
 
 // SetMeta sets given metadata to this document.
 func (n *Document) SetMeta(meta map[string]interface{}) {
-	n.meta = meta
+	if n.meta == nil {
+		n.meta = map[string]interface{}{}
+	}
+	for k, v := range meta {
+		n.meta[k] = v
+	}
+}
+
+// AddMeta adds given metadata to this document.
+func (n *Document) AddMeta(key string, value interface{}) {
+	if n.meta == nil {
+		n.meta = map[string]interface{}{}
+	}
+	n.meta[key] = value
 }
 
 // NewDocument returns a new Document node.

vendor/github.com/yuin/goldmark/extension/typographer.go

@@ -160,7 +160,7 @@ func NewTypographerParser(opts ...TypographerOption) parser.InlineParser {
 }
 
 func (s *typographerParser) Trigger() []byte {
-	return []byte{'\'', '"', '-', '.', '<', '>'}
+	return []byte{'\'', '"', '-', '.', ',', '<', '>', '*', '['}
 }
 
 func (s *typographerParser) Parse(parent gast.Node, block text.Reader, pc parser.Context) gast.Node {
@@ -230,6 +230,13 @@ func (s *typographerParser) Parse(parent gast.Node, block text.Reader, pc parser
 						return node
 					}
 				}
+				// special cases: 'twas, 'em, 'net
+				if len(line) > 1 && (unicode.IsPunct(before) || unicode.IsSpace(before)) && (line[1] == 't' || line[1] == 'e' || line[1] == 'n' || line[1] == 'l') {
+					node := gast.NewString(s.Substitutions[Apostrophe])
+					node.SetCode(true)
+					block.Advance(1)
+					return node
+				}
 				// Convert normal apostrophes. This is probably more flexible than necessary but
 				// converts any apostrophe in between two alphanumerics.
 				if len(line) > 1 && (unicode.IsDigit(before) || unicode.IsLetter(before)) && (unicode.IsLetter(util.ToRune(line, 1))) {
@@ -241,7 +248,7 @@ func (s *typographerParser) Parse(parent gast.Node, block text.Reader, pc parser
 			}
 			if s.Substitutions[LeftSingleQuote] != nil && d.CanOpen && !d.CanClose {
 				nt := LeftSingleQuote
-				// special cases: Alice's, I'm ,Don't, You'd
+				// special cases: Alice's, I'm, Don't, You'd
 				if len(line) > 1 && (line[1] == 's' || line[1] == 'm' || line[1] == 't' || line[1] == 'd') && (len(line) < 3 || util.IsPunct(line[2]) || util.IsSpace(line[2])) {
 					nt = RightSingleQuote
 				}
@@ -258,9 +265,18 @@ func (s *typographerParser) Parse(parent gast.Node, block text.Reader, pc parser
 				block.Advance(1)
 				return node
 			}
+			if s.Substitutions[RightSingleQuote] != nil {
+				// plural possesives and abbreviations: Smiths', doin'
+				if len(line) > 1 && unicode.IsSpace(util.ToRune(line, 0)) || unicode.IsPunct(util.ToRune(line, 0)) && (len(line) > 2 && !unicode.IsDigit(util.ToRune(line, 1))) {
+					node := gast.NewString(s.Substitutions[RightSingleQuote])
+					node.SetCode(true)
+					block.Advance(1)
+					return node
+				}
+			}
 			if s.Substitutions[RightSingleQuote] != nil && counter.Single > 0 {
 				isClose := d.CanClose && !d.CanOpen
-				maybeClose := d.CanClose && d.CanOpen && len(line) > 1 && (line[1] == ',' || line[1] == '.' || line[1] == '!' || line[1] == '?') && (len(line) == 2 || (len(line) > 2 && util.IsPunct(line[2]) || util.IsSpace(line[2])))
+				maybeClose := d.CanClose && d.CanOpen && len(line) > 1 && unicode.IsPunct(util.ToRune(line, 1)) && (len(line) == 2 || (len(line) > 2 && util.IsPunct(line[2]) || util.IsSpace(line[2])))
 				if isClose || maybeClose {
 					node := gast.NewString(s.Substitutions[RightSingleQuote])
 					node.SetCode(true)
@@ -280,7 +296,7 @@ func (s *typographerParser) Parse(parent gast.Node, block text.Reader, pc parser
 			}
 			if s.Substitutions[RightDoubleQuote] != nil && counter.Double > 0 {
 				isClose := d.CanClose && !d.CanOpen
-				maybeClose := d.CanClose && d.CanOpen && len(line) > 1 && (line[1] == ',' || line[1] == '.' || line[1] == '!' || line[1] == '?') && (len(line) == 2 || (len(line) > 2 && util.IsPunct(line[2]) || util.IsSpace(line[2])))
+				maybeClose := d.CanClose && d.CanOpen && len(line) > 1 && (unicode.IsPunct(util.ToRune(line, 1))) && (len(line) == 2 || (len(line) > 2 && util.IsPunct(line[2]) || util.IsSpace(line[2])))
 				if isClose || maybeClose {
 					// special case: "Monitor 21""
 					if len(line) > 1 && line[1] == '"' && unicode.IsDigit(before) {