Bump github.com/klauspost/compress from 1.9.1 to 1.9.2

Bumps [github.com/klauspost/compress](https://github.com/klauspost/compress) from 1.9.1 to 1.9.2.
- [Release notes](https://github.com/klauspost/compress/releases)
- [Changelog](https://github.com/klauspost/compress/blob/master/.goreleaser.yml)
- [Commits](https://github.com/klauspost/compress/compare/v1.9.1...v1.9.2)

Signed-off-by: dependabot-preview[bot] <support@dependabot.com>
Signed-off-by: Daniel J Walsh <dwalsh@redhat.com>

go.mod

@@ -7,7 +7,7 @@ require (
 	github.com/Microsoft/hcsshim v0.8.6
 	github.com/docker/docker v0.0.0-20171019062838-86f080cff091 // indirect
 	github.com/docker/go-units v0.4.0
-	github.com/klauspost/compress v1.9.1
+	github.com/klauspost/compress v1.9.2
 	github.com/klauspost/cpuid v1.2.1 // indirect
 	github.com/klauspost/pgzip v1.2.1
 	github.com/mattn/go-shellwords v1.0.6

go.sum

@@ -35,6 +35,8 @@ github.com/klauspost/compress v1.7.2 h1:liMOoeIvFpr9kEvalrZ7VVBA4wGf7zfOgwBjzz/5
 github.com/klauspost/compress v1.7.2/go.mod h1:RyIbtBH6LamlWaDj8nUwkbUhJ87Yi3uG0guNDohfE1A=
 github.com/klauspost/compress v1.9.1 h1:TWy0o9J9c6LK9C8t7Msh6IAJNXbsU/nvKLTQUU5HdaY=
 github.com/klauspost/compress v1.9.1/go.mod h1:RyIbtBH6LamlWaDj8nUwkbUhJ87Yi3uG0guNDohfE1A=
+github.com/klauspost/compress v1.9.2 h1:LfVyl+ZlLlLDeQ/d2AqfGIIH4qEDu0Ed2S5GyhCWIWY=
+github.com/klauspost/compress v1.9.2/go.mod h1:RyIbtBH6LamlWaDj8nUwkbUhJ87Yi3uG0guNDohfE1A=
 github.com/klauspost/cpuid v1.2.1 h1:vJi+O/nMdFt0vqm8NZBI6wzALWdA2X+egi0ogNyrC/w=
 github.com/klauspost/cpuid v1.2.1/go.mod h1:Pj4uuM528wm8OyEC2QMXAi2YiTZ96dNQPGgoMS4s3ek=
 github.com/klauspost/pgzip v1.2.1 h1:oIPZROsWuPHpOdMVWLuJZXwgjhrW8r1yEX8UqMyeNHM=

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

@@ -322,8 +322,6 @@ func (d *compressor) writeStoredBlock(buf []byte) error {
 	return d.w.err
 }
 
-const hashmul = 0x1e35a7bd
-
 // hash4 returns a hash representation of the first 4 bytes
 // of the supplied slice.
 // The caller must ensure that len(b) >= 4.

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

@@ -87,9 +87,6 @@ type huffmanBitWriter struct {
 	bits            uint64
 	nbits           uint16
 	nbytes          uint8
-	literalFreq     []uint16
-	offsetFreq      []uint16
-	codegen         []uint8
 	literalEncoding *huffmanEncoder
 	offsetEncoding  *huffmanEncoder
 	codegenEncoding *huffmanEncoder
@@ -99,7 +96,12 @@ type huffmanBitWriter struct {
 	logReusePenalty uint
 	lastHuffMan     bool
 	bytes           [256]byte
+	literalFreq     [lengthCodesStart + 32]uint16
+	offsetFreq      [32]uint16
 	codegenFreq     [codegenCodeCount]uint16
+
+	// codegen must have an extra space for the final symbol.
+	codegen [literalCount + offsetCodeCount + 1]uint8
 }
 
 // Huffman reuse.
@@ -124,10 +126,6 @@ type huffmanBitWriter struct {
 func newHuffmanBitWriter(w io.Writer) *huffmanBitWriter {
 	return &huffmanBitWriter{
 		writer:          w,
-		literalFreq: make([]uint16, lengthCodesStart+32),
-		offsetFreq:  make([]uint16, 32),
-		// codegen must have an extra space for the final symbol.
-		codegen:         make([]uint8, literalCount+offsetCodeCount+1),
 		literalEncoding: newHuffmanEncoder(literalCount),
 		codegenEncoding: newHuffmanEncoder(codegenCodeCount),
 		offsetEncoding:  newHuffmanEncoder(offsetCodeCount),
@@ -253,7 +251,7 @@ func (w *huffmanBitWriter) generateCodegen(numLiterals int, numOffsets int, litE
 	// a copy of the frequencies, and as the place where we put the result.
 	// This is fine because the output is always shorter than the input used
 	// so far.
-	codegen := w.codegen // cache
+	codegen := w.codegen[:] // cache
 	// Copy the concatenated code sizes to codegen. Put a marker at the end.
 	cgnl := codegen[:numLiterals]
 	for i := range cgnl {
@@ -356,8 +354,8 @@ func (w *huffmanBitWriter) headerSize() (size, numCodegens int) {
 func (w *huffmanBitWriter) dynamicSize(litEnc, offEnc *huffmanEncoder, extraBits int) (size, numCodegens int) {
 	header, numCodegens := w.headerSize()
 	size = header +
-		litEnc.bitLength(w.literalFreq) +
+		litEnc.bitLength(w.literalFreq[:]) +
-		offEnc.bitLength(w.offsetFreq) +
+		offEnc.bitLength(w.offsetFreq[:]) +
 		extraBits
 	return size, numCodegens
 }
@@ -378,8 +376,8 @@ func (w *huffmanBitWriter) extraBitSize() int {
 // fixedSize returns the size of dynamically encoded data in bits.
 func (w *huffmanBitWriter) fixedSize(extraBits int) int {
 	return 3 +
-		fixedLiteralEncoding.bitLength(w.literalFreq) +
+		fixedLiteralEncoding.bitLength(w.literalFreq[:]) +
-		fixedOffsetEncoding.bitLength(w.offsetFreq) +
+		fixedOffsetEncoding.bitLength(w.offsetFreq[:]) +
 		extraBits
 }
 
@@ -670,9 +668,9 @@ func (w *huffmanBitWriter) writeBlockDynamic(tokens *tokens, eof bool, input []b
 // and offsetEncoding.
 // The number of literal and offset tokens is returned.
 func (w *huffmanBitWriter) indexTokens(t *tokens, filled bool) (numLiterals, numOffsets int) {
-	copy(w.literalFreq, t.litHist[:])
+	copy(w.literalFreq[:], t.litHist[:])
 	copy(w.literalFreq[256:], t.extraHist[:])
-	copy(w.offsetFreq, t.offHist[:offsetCodeCount])
+	copy(w.offsetFreq[:], t.offHist[:offsetCodeCount])
 
 	if t.n == 0 {
 		return
@@ -797,17 +795,17 @@ func (w *huffmanBitWriter) writeBlockHuff(eof bool, input []byte, sync bool) {
 	}
 
 	// Clear histogram
-	for i := range w.literalFreq {
+	for i := range w.literalFreq[:] {
 		w.literalFreq[i] = 0
 	}
 	if !w.lastHuffMan {
-		for i := range w.offsetFreq {
+		for i := range w.offsetFreq[:] {
 			w.offsetFreq[i] = 0
 		}
 	}
 
 	// Add everything as literals
-	estBits := histogramSize(input, w.literalFreq, !eof && !sync) + 15
+	estBits := histogramSize(input, w.literalFreq[:], !eof && !sync) + 15
 
 	// Store bytes, if we don't get a reasonable improvement.
 	ssize, storable := w.storedSize(input)

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

@@ -107,8 +107,8 @@ const (
 
 type huffmanDecoder struct {
 	min      int                       // the minimum code length
-	chunks   *[huffmanNumChunks]uint32 // chunks as described above
+	chunks   *[huffmanNumChunks]uint16 // chunks as described above
-	links    [][]uint32                // overflow links
+	links    [][]uint16                // overflow links
 	linkMask uint32                    // mask the width of the link table
 }
 
@@ -124,7 +124,7 @@ func (h *huffmanDecoder) init(lengths []int) bool {
 	const sanity = false
 
 	if h.chunks == nil {
-		h.chunks = &[huffmanNumChunks]uint32{}
+		h.chunks = &[huffmanNumChunks]uint16{}
 	}
 	if h.min != 0 {
 		*h = huffmanDecoder{chunks: h.chunks, links: h.links}
@@ -191,7 +191,7 @@ func (h *huffmanDecoder) init(lengths []int) bool {
 		// create link tables
 		link := nextcode[huffmanChunkBits+1] >> 1
 		if cap(h.links) < huffmanNumChunks-link {
-			h.links = make([][]uint32, huffmanNumChunks-link)
+			h.links = make([][]uint16, huffmanNumChunks-link)
 		} else {
 			h.links = h.links[:huffmanNumChunks-link]
 		}
@@ -202,9 +202,9 @@ func (h *huffmanDecoder) init(lengths []int) bool {
 			if sanity && h.chunks[reverse] != 0 {
 				panic("impossible: overwriting existing chunk")
 			}
-			h.chunks[reverse] = uint32(off<<huffmanValueShift | (huffmanChunkBits + 1))
+			h.chunks[reverse] = uint16(off<<huffmanValueShift | (huffmanChunkBits + 1))
 			if cap(h.links[off]) < numLinks {
-				h.links[off] = make([]uint32, numLinks)
+				h.links[off] = make([]uint16, numLinks)
 			} else {
 				links := h.links[off][:0]
 				h.links[off] = links[:numLinks]
@@ -220,7 +220,7 @@ func (h *huffmanDecoder) init(lengths []int) bool {
 		}
 		code := nextcode[n]
 		nextcode[n]++
-		chunk := uint32(i<<huffmanValueShift | n)
+		chunk := uint16(i<<huffmanValueShift | n)
 		reverse := int(bits.Reverse16(uint16(code)))
 		reverse >>= uint(16 - n)
 		if n <= huffmanChunkBits {

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

@@ -0,0 +1,252 @@
+package flate
+
+import (
+	"io"
+	"math"
+)
+
+const (
+	maxStatelessBlock = math.MaxInt16
+
+	slTableBits  = 13
+	slTableSize  = 1 << slTableBits
+	slTableShift = 32 - slTableBits
+)
+
+type statelessWriter struct {
+	dst    io.Writer
+	closed bool
+}
+
+func (s *statelessWriter) Close() error {
+	if s.closed {
+		return nil
+	}
+	s.closed = true
+	// Emit EOF block
+	return StatelessDeflate(s.dst, nil, true)
+}
+
+func (s *statelessWriter) Write(p []byte) (n int, err error) {
+	err = StatelessDeflate(s.dst, p, false)
+	if err != nil {
+		return 0, err
+	}
+	return len(p), nil
+}
+
+func (s *statelessWriter) Reset(w io.Writer) {
+	s.dst = w
+	s.closed = false
+}
+
+// NewStatelessWriter will do compression but without maintaining any state
+// between Write calls.
+// There will be no memory kept between Write calls,
+// but compression and speed will be suboptimal.
+// Because of this, the size of actual Write calls will affect output size.
+func NewStatelessWriter(dst io.Writer) io.WriteCloser {
+	return &statelessWriter{dst: dst}
+}
+
+// StatelessDeflate allows to compress directly to a Writer without retaining state.
+// When returning everything will be flushed.
+func StatelessDeflate(out io.Writer, in []byte, eof bool) error {
+	var dst tokens
+	bw := newHuffmanBitWriter(out)
+	if eof && len(in) == 0 {
+		// Just write an EOF block.
+		// Could be faster...
+		bw.writeStoredHeader(0, true)
+		bw.flush()
+		return bw.err
+	}
+
+	for len(in) > 0 {
+		todo := in
+		if len(todo) > maxStatelessBlock {
+			todo = todo[:maxStatelessBlock]
+		}
+		in = in[len(todo):]
+		// Compress
+		statelessEnc(&dst, todo)
+		isEof := eof && len(in) == 0
+
+		if dst.n == 0 {
+			bw.writeStoredHeader(len(todo), isEof)
+			if bw.err != nil {
+				return bw.err
+			}
+			bw.writeBytes(todo)
+		} else if int(dst.n) > len(todo)-len(todo)>>4 {
+			// If we removed less than 1/16th, huffman compress the block.
+			bw.writeBlockHuff(isEof, todo, false)
+		} else {
+			bw.writeBlockDynamic(&dst, isEof, todo, false)
+		}
+		if bw.err != nil {
+			return bw.err
+		}
+		dst.Reset()
+	}
+	if !eof {
+		// Align.
+		bw.writeStoredHeader(0, false)
+	}
+	bw.flush()
+	return bw.err
+}
+
+func hashSL(u uint32) uint32 {
+	return (u * 0x1e35a7bd) >> slTableShift
+}
+
+func load3216(b []byte, i int16) uint32 {
+	// Help the compiler eliminate bounds checks on the read so it can be done in a single read.
+	b = b[i:]
+	b = b[:4]
+	return uint32(b[0]) | uint32(b[1])<<8 | uint32(b[2])<<16 | uint32(b[3])<<24
+}
+
+func load6416(b []byte, i int16) uint64 {
+	// Help the compiler eliminate bounds checks on the read so it can be done in a single read.
+	b = b[i:]
+	b = b[:8]
+	return uint64(b[0]) | uint64(b[1])<<8 | uint64(b[2])<<16 | uint64(b[3])<<24 |
+		uint64(b[4])<<32 | uint64(b[5])<<40 | uint64(b[6])<<48 | uint64(b[7])<<56
+}
+
+func statelessEnc(dst *tokens, src []byte) {
+	const (
+		inputMargin            = 12 - 1
+		minNonLiteralBlockSize = 1 + 1 + inputMargin
+	)
+
+	type tableEntry struct {
+		offset int16
+	}
+
+	var table [slTableSize]tableEntry
+
+	// This check isn't in the Snappy implementation, but there, the caller
+	// instead of the callee handles this case.
+	if len(src) < minNonLiteralBlockSize {
+		// We do not fill the token table.
+		// This will be picked up by caller.
+		dst.n = uint16(len(src))
+		return
+	}
+
+	s := int16(1)
+	nextEmit := int16(0)
+	// sLimit is when to stop looking for offset/length copies. The inputMargin
+	// lets us use a fast path for emitLiteral in the main loop, while we are
+	// looking for copies.
+	sLimit := int16(len(src) - inputMargin)
+
+	// nextEmit is where in src the next emitLiteral should start from.
+	cv := load3216(src, s)
+
+	for {
+		const skipLog = 5
+		const doEvery = 2
+
+		nextS := s
+		var candidate tableEntry
+		for {
+			nextHash := hashSL(cv)
+			candidate = table[nextHash]
+			nextS = s + doEvery + (s-nextEmit)>>skipLog
+			if nextS > sLimit || nextS <= 0 {
+				goto emitRemainder
+			}
+
+			now := load6416(src, nextS)
+			table[nextHash] = tableEntry{offset: s}
+			nextHash = hashSL(uint32(now))
+
+			if cv == load3216(src, candidate.offset) {
+				table[nextHash] = tableEntry{offset: nextS}
+				break
+			}
+
+			// Do one right away...
+			cv = uint32(now)
+			s = nextS
+			nextS++
+			candidate = table[nextHash]
+			now >>= 8
+			table[nextHash] = tableEntry{offset: s}
+
+			if cv == load3216(src, candidate.offset) {
+				table[nextHash] = tableEntry{offset: nextS}
+				break
+			}
+			cv = uint32(now)
+			s = nextS
+		}
+
+		// A 4-byte match has been found. We'll later see if more than 4 bytes
+		// match. But, prior to the match, src[nextEmit:s] are unmatched. Emit
+		// them as literal bytes.
+		for {
+			// Invariant: we have a 4-byte match at s, and no need to emit any
+			// literal bytes prior to s.
+
+			// Extend the 4-byte match as long as possible.
+			t := candidate.offset
+			l := int16(matchLen(src[s+4:], src[t+4:]) + 4)
+
+			// Extend backwards
+			for t > 0 && s > nextEmit && src[t-1] == src[s-1] {
+				s--
+				t--
+				l++
+			}
+			if nextEmit < s {
+				emitLiteral(dst, src[nextEmit:s])
+			}
+
+			// Save the match found
+			dst.AddMatchLong(int32(l), uint32(s-t-baseMatchOffset))
+			s += l
+			nextEmit = s
+			if nextS >= s {
+				s = nextS + 1
+			}
+			if s >= sLimit {
+				goto emitRemainder
+			}
+
+			// We could immediately start working at s now, but to improve
+			// compression we first update the hash table at s-2 and at s. If
+			// another emitCopy is not our next move, also calculate nextHash
+			// at s+1. At least on GOARCH=amd64, these three hash calculations
+			// are faster as one load64 call (with some shifts) instead of
+			// three load32 calls.
+			x := load6416(src, s-2)
+			o := s - 2
+			prevHash := hashSL(uint32(x))
+			table[prevHash] = tableEntry{offset: o}
+			x >>= 16
+			currHash := hashSL(uint32(x))
+			candidate = table[currHash]
+			table[currHash] = tableEntry{offset: o + 2}
+
+			if uint32(x) != load3216(src, candidate.offset) {
+				cv = uint32(x >> 8)
+				s++
+				break
+			}
+		}
+	}
+
+emitRemainder:
+	if int(nextEmit) < len(src) {
+		// If nothing was added, don't encode literals.
+		if dst.n == 0 {
+			return
+		}
+		emitLiteral(dst, src[nextEmit:])
+	}
+}

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

@@ -161,7 +161,8 @@ func (b *blockDec) reset(br byteBuffer, windowSize uint64) error {
 	b.data, err = br.readBig(cSize, b.dataStorage)
 	if err != nil {
 		if debug {
-			println("Reading block:", err)
+			println("Reading block:", err, "(", cSize, ")", len(b.data))
+			printf("%T", br)
 		}
 		return err
 	}

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

@@ -101,6 +101,9 @@ func (r *readerWrapper) readBig(n int, dst []byte) ([]byte, error) {
 		dst = make([]byte, n)
 	}
 	n2, err := io.ReadFull(r.r, dst[:n])
+	if err == io.EOF && n > 0 {
+		err = io.ErrUnexpectedEOF
+	}
 	return dst[:n2], err
 }
 

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

@@ -194,14 +194,14 @@ func (d *frameDec) reset(br byteBuffer) error {
 			// When FCS_Field_Size is 2, the offset of 256 is added.
 			d.FrameContentSize = uint64(b[0]) | (uint64(b[1]) << 8) + 256
 		case 4:
-			d.FrameContentSize = uint64(b[0]) | (uint64(b[1]) << 8) | (uint64(b[2]) << 16) | (uint64(b[3] << 24))
+			d.FrameContentSize = uint64(b[0]) | (uint64(b[1]) << 8) | (uint64(b[2]) << 16) | (uint64(b[3]) << 24)
 		case 8:
 			d1 := uint32(b[0]) | (uint32(b[1]) << 8) | (uint32(b[2]) << 16) | (uint32(b[3]) << 24)
 			d2 := uint32(b[4]) | (uint32(b[5]) << 8) | (uint32(b[6]) << 16) | (uint32(b[7]) << 24)
 			d.FrameContentSize = uint64(d1) | (uint64(d2) << 32)
 		}
 		if debug {
-			println("field size bits:", v, "fcsSize:", fcsSize, "FrameContentSize:", d.FrameContentSize, hex.EncodeToString(b[:fcsSize]))
+			println("field size bits:", v, "fcsSize:", fcsSize, "FrameContentSize:", d.FrameContentSize, hex.EncodeToString(b[:fcsSize]), "singleseg:", d.SingleSegment, "window:", d.WindowSize)
 		}
 	}
 	// Move this to shared.
@@ -414,6 +414,7 @@ func (d *frameDec) startDecoder(output chan decodeOutput) {
 		}
 		written += int64(len(r.b))
 		if d.SingleSegment && uint64(written) > d.FrameContentSize {
+			println("runDecoder: single segment and", uint64(written), ">", d.FrameContentSize)
 			r.err = ErrFrameSizeExceeded
 			output <- r
 			return
@@ -464,6 +465,7 @@ func (d *frameDec) runDecoder(dst []byte, dec *blockDec) ([]byte, error) {
 			break
 		}
 		if d.SingleSegment && uint64(len(d.history.b)) > d.o.maxDecodedSize {
+			println("runDecoder: single segment and", uint64(len(d.history.b)), ">", d.o.maxDecodedSize)
 			err = ErrFrameSizeExceeded
 			break
 		}

vendor/modules.txt

@@ -30,7 +30,7 @@ github.com/google/go-cmp/cmp
 github.com/google/go-cmp/cmp/internal/diff
 github.com/google/go-cmp/cmp/internal/function
 github.com/google/go-cmp/cmp/internal/value
-# github.com/klauspost/compress v1.9.1
+# github.com/klauspost/compress v1.9.2
 github.com/klauspost/compress/flate
 github.com/klauspost/compress/fse
 github.com/klauspost/compress/huff0