Use goreleaser to build and release (#244)

Instead of using gox on one side and an action to release, we can merge
them together with goreleaser which will build for extra targets (arm,
mips if needed in the future) and it also takes care of creating
checksums, a source archive, and a changelog and creating a release with
all the artifacts.

All binaries should respect the old naming convention, so any scripts
out there should still work.

Signed-off-by: Itxaka <igarcia@suse.com>

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

@@ -62,8 +62,10 @@ type sequenceDecs struct {
 	matchLengths sequenceDec
 	prevOffset   [3]int
 	hist         []byte
+	dict         []byte
 	literals     []byte
 	out          []byte
+	windowSize   int
 	maxBits      uint8
 }
 
@@ -82,7 +84,12 @@ func (s *sequenceDecs) initialize(br *bitReader, hist *history, literals, out []
 	s.hist = hist.b
 	s.prevOffset = hist.recentOffsets
 	s.maxBits = s.litLengths.fse.maxBits + s.offsets.fse.maxBits + s.matchLengths.fse.maxBits
+	s.windowSize = hist.windowSize
 	s.out = out
+	s.dict = nil
+	if hist.dict != nil {
+		s.dict = hist.dict.content
+	}
 	return nil
 }
 
@@ -98,76 +105,159 @@ func (s *sequenceDecs) decode(seqs int, br *bitReader, hist []byte) error {
 			printf("reading sequence %d, exceeded available data\n", seqs-i)
 			return io.ErrUnexpectedEOF
 		}
-		var litLen, matchOff, matchLen int
+		var ll, mo, ml int
 		if br.off > 4+((maxOffsetBits+16+16)>>3) {
-			litLen, matchOff, matchLen = s.nextFast(br, llState, mlState, ofState)
+			// inlined function:
+			// ll, mo, ml = s.nextFast(br, llState, mlState, ofState)
+
+			// Final will not read from stream.
+			var llB, mlB, moB uint8
+			ll, llB = llState.final()
+			ml, mlB = mlState.final()
+			mo, moB = ofState.final()
+
+			// extra bits are stored in reverse order.
+			br.fillFast()
+			mo += br.getBits(moB)
+			if s.maxBits > 32 {
+				br.fillFast()
+			}
+			ml += br.getBits(mlB)
+			ll += br.getBits(llB)
+
+			if moB > 1 {
+				s.prevOffset[2] = s.prevOffset[1]
+				s.prevOffset[1] = s.prevOffset[0]
+				s.prevOffset[0] = mo
+			} else {
+				// mo = s.adjustOffset(mo, ll, moB)
+				// Inlined for rather big speedup
+				if ll == 0 {
+					// There is an exception though, when current sequence's literals_length = 0.
+					// In this case, repeated offsets are shifted by one, so an offset_value of 1 means Repeated_Offset2,
+					// an offset_value of 2 means Repeated_Offset3, and an offset_value of 3 means Repeated_Offset1 - 1_byte.
+					mo++
+				}
+
+				if mo == 0 {
+					mo = s.prevOffset[0]
+				} else {
+					var temp int
+					if mo == 3 {
+						temp = s.prevOffset[0] - 1
+					} else {
+						temp = s.prevOffset[mo]
+					}
+
+					if temp == 0 {
+						// 0 is not valid; input is corrupted; force offset to 1
+						println("temp was 0")
+						temp = 1
+					}
+
+					if mo != 1 {
+						s.prevOffset[2] = s.prevOffset[1]
+					}
+					s.prevOffset[1] = s.prevOffset[0]
+					s.prevOffset[0] = temp
+					mo = temp
+				}
+			}
 			br.fillFast()
 		} else {
-			litLen, matchOff, matchLen = s.next(br, llState, mlState, ofState)
+			ll, mo, ml = s.next(br, llState, mlState, ofState)
 			br.fill()
 		}
 
 		if debugSequences {
-			println("Seq", seqs-i-1, "Litlen:", litLen, "matchOff:", matchOff, "(abs) matchLen:", matchLen)
+			println("Seq", seqs-i-1, "Litlen:", ll, "mo:", mo, "(abs) ml:", ml)
 		}
 
-		if litLen > len(s.literals) {
-			return fmt.Errorf("unexpected literal count, want %d bytes, but only %d is available", litLen, len(s.literals))
+		if ll > len(s.literals) {
+			return fmt.Errorf("unexpected literal count, want %d bytes, but only %d is available", ll, len(s.literals))
 		}
-		size := litLen + matchLen + len(s.out)
+		size := ll + ml + len(s.out)
 		if size-startSize > maxBlockSize {
 			return fmt.Errorf("output (%d) bigger than max block size", size)
 		}
 		if size > cap(s.out) {
-			// Not enough size, will be extremely rarely triggered,
+			// Not enough size, which can happen under high volume block streaming conditions
 			// but could be if destination slice is too small for sync operations.
-			// We add maxBlockSize to the capacity.
-			s.out = append(s.out, make([]byte, maxBlockSize)...)
-			s.out = s.out[:len(s.out)-maxBlockSize]
+			// over-allocating here can create a large amount of GC pressure so we try to keep
+			// it as contained as possible
+			used := len(s.out) - startSize
+			addBytes := 256 + ll + ml + used>>2
+			// Clamp to max block size.
+			if used+addBytes > maxBlockSize {
+				addBytes = maxBlockSize - used
+			}
+			s.out = append(s.out, make([]byte, addBytes)...)
+			s.out = s.out[:len(s.out)-addBytes]
 		}
-		if matchLen > maxMatchLen {
-			return fmt.Errorf("match len (%d) bigger than max allowed length", matchLen)
-		}
-		if matchOff > len(s.out)+len(hist)+litLen {
-			return fmt.Errorf("match offset (%d) bigger than current history (%d)", matchOff, len(s.out)+len(hist)+litLen)
-		}
-		if matchOff == 0 && matchLen > 0 {
-			return fmt.Errorf("zero matchoff and matchlen > 0")
+		if ml > maxMatchLen {
+			return fmt.Errorf("match len (%d) bigger than max allowed length", ml)
 		}
 
-		s.out = append(s.out, s.literals[:litLen]...)
-		s.literals = s.literals[litLen:]
+		// Add literals
+		s.out = append(s.out, s.literals[:ll]...)
+		s.literals = s.literals[ll:]
 		out := s.out
 
+		if mo == 0 && ml > 0 {
+			return fmt.Errorf("zero matchoff and matchlen (%d) > 0", ml)
+		}
+
+		if mo > len(s.out)+len(hist) || mo > s.windowSize {
+			if len(s.dict) == 0 {
+				return fmt.Errorf("match offset (%d) bigger than current history (%d)", mo, len(s.out)+len(hist))
+			}
+
+			// we may be in dictionary.
+			dictO := len(s.dict) - (mo - (len(s.out) + len(hist)))
+			if dictO < 0 || dictO >= len(s.dict) {
+				return fmt.Errorf("match offset (%d) bigger than current history (%d)", mo, len(s.out)+len(hist))
+			}
+			end := dictO + ml
+			if end > len(s.dict) {
+				out = append(out, s.dict[dictO:]...)
+				mo -= len(s.dict) - dictO
+				ml -= len(s.dict) - dictO
+			} else {
+				out = append(out, s.dict[dictO:end]...)
+				mo = 0
+				ml = 0
+			}
+		}
+
 		// Copy from history.
 		// TODO: Blocks without history could be made to ignore this completely.
-		if v := matchOff - len(s.out); v > 0 {
+		if v := mo - len(s.out); v > 0 {
 			// v is the start position in history from end.
 			start := len(s.hist) - v
-			if matchLen > v {
+			if ml > v {
 				// Some goes into current block.
 				// Copy remainder of history
 				out = append(out, s.hist[start:]...)
-				matchOff -= v
-				matchLen -= v
+				mo -= v
+				ml -= v
 			} else {
-				out = append(out, s.hist[start:start+matchLen]...)
-				matchLen = 0
+				out = append(out, s.hist[start:start+ml]...)
+				ml = 0
 			}
 		}
 		// We must be in current buffer now
-		if matchLen > 0 {
-			start := len(s.out) - matchOff
-			if matchLen <= len(s.out)-start {
+		if ml > 0 {
+			start := len(s.out) - mo
+			if ml <= len(s.out)-start {
 				// No overlap
-				out = append(out, s.out[start:start+matchLen]...)
+				out = append(out, s.out[start:start+ml]...)
 			} else {
 				// Overlapping copy
 				// Extend destination slice and copy one byte at the time.
-				out = out[:len(out)+matchLen]
-				src := out[start : start+matchLen]
+				out = out[:len(out)+ml]
+				src := out[start : start+ml]
 				// Destination is the space we just added.
-				dst := out[len(out)-matchLen:]
+				dst := out[len(out)-ml:]
 				dst = dst[:len(src)]
 				for i := range src {
 					dst[i] = src[i]