mirror of
https://github.com/go-gitea/gitea.git
synced 2024-12-15 17:49:20 +08:00
12a1f914f4
* update github.com/alecthomas/chroma v0.8.0 -> v0.8.1 * github.com/blevesearch/bleve v1.0.10 -> v1.0.12 * editorconfig-core-go v2.1.1 -> v2.3.7 * github.com/gliderlabs/ssh v0.2.2 -> v0.3.1 * migrate editorconfig.ParseBytes to Parse * github.com/shurcooL/vfsgen to 0d455de96546 * github.com/go-git/go-git/v5 v5.1.0 -> v5.2.0 * github.com/google/uuid v1.1.1 -> v1.1.2 * github.com/huandu/xstrings v1.3.0 -> v1.3.2 * github.com/klauspost/compress v1.10.11 -> v1.11.1 * github.com/markbates/goth v1.61.2 -> v1.65.0 * github.com/mattn/go-sqlite3 v1.14.0 -> v1.14.4 * github.com/mholt/archiver v3.3.0 -> v3.3.2 * github.com/microcosm-cc/bluemonday 4f7140c49acb -> v1.0.4 * github.com/minio/minio-go v7.0.4 -> v7.0.5 * github.com/olivere/elastic v7.0.9 -> v7.0.20 * github.com/urfave/cli v1.20.0 -> v1.22.4 * github.com/prometheus/client_golang v1.1.0 -> v1.8.0 * github.com/xanzy/go-gitlab v0.37.0 -> v0.38.1 * mvdan.cc/xurls v2.1.0 -> v2.2.0 Co-authored-by: Lauris BH <lauris@nix.lv>
399 lines
8.1 KiB
Go
Vendored
399 lines
8.1 KiB
Go
Vendored
// Copyright 2014-2019 Ulrich Kunitz. All rights reserved.
|
|
// Use of this source code is governed by a BSD-style
|
|
// license that can be found in the LICENSE file.
|
|
|
|
package lzma
|
|
|
|
import (
|
|
"errors"
|
|
"fmt"
|
|
"io"
|
|
)
|
|
|
|
const (
|
|
// maximum size of compressed data in a chunk
|
|
maxCompressed = 1 << 16
|
|
// maximum size of uncompressed data in a chunk
|
|
maxUncompressed = 1 << 21
|
|
)
|
|
|
|
// chunkType represents the type of an LZMA2 chunk. Note that this
|
|
// value is an internal representation and no actual encoding of a LZMA2
|
|
// chunk header.
|
|
type chunkType byte
|
|
|
|
// Possible values for the chunk type.
|
|
const (
|
|
// end of stream
|
|
cEOS chunkType = iota
|
|
// uncompressed; reset dictionary
|
|
cUD
|
|
// uncompressed; no reset of dictionary
|
|
cU
|
|
// LZMA compressed; no reset
|
|
cL
|
|
// LZMA compressed; reset state
|
|
cLR
|
|
// LZMA compressed; reset state; new property value
|
|
cLRN
|
|
// LZMA compressed; reset state; new property value; reset dictionary
|
|
cLRND
|
|
)
|
|
|
|
// chunkTypeStrings provide a string representation for the chunk types.
|
|
var chunkTypeStrings = [...]string{
|
|
cEOS: "EOS",
|
|
cU: "U",
|
|
cUD: "UD",
|
|
cL: "L",
|
|
cLR: "LR",
|
|
cLRN: "LRN",
|
|
cLRND: "LRND",
|
|
}
|
|
|
|
// String returns a string representation of the chunk type.
|
|
func (c chunkType) String() string {
|
|
if !(cEOS <= c && c <= cLRND) {
|
|
return "unknown"
|
|
}
|
|
return chunkTypeStrings[c]
|
|
}
|
|
|
|
// Actual encodings for the chunk types in the value. Note that the high
|
|
// uncompressed size bits are stored in the header byte additionally.
|
|
const (
|
|
hEOS = 0
|
|
hUD = 1
|
|
hU = 2
|
|
hL = 1 << 7
|
|
hLR = 1<<7 | 1<<5
|
|
hLRN = 1<<7 | 1<<6
|
|
hLRND = 1<<7 | 1<<6 | 1<<5
|
|
)
|
|
|
|
// errHeaderByte indicates an unsupported value for the chunk header
|
|
// byte. These bytes starts the variable-length chunk header.
|
|
var errHeaderByte = errors.New("lzma: unsupported chunk header byte")
|
|
|
|
// headerChunkType converts the header byte into a chunk type. It
|
|
// ignores the uncompressed size bits in the chunk header byte.
|
|
func headerChunkType(h byte) (c chunkType, err error) {
|
|
if h&hL == 0 {
|
|
// no compression
|
|
switch h {
|
|
case hEOS:
|
|
c = cEOS
|
|
case hUD:
|
|
c = cUD
|
|
case hU:
|
|
c = cU
|
|
default:
|
|
return 0, errHeaderByte
|
|
}
|
|
return
|
|
}
|
|
switch h & hLRND {
|
|
case hL:
|
|
c = cL
|
|
case hLR:
|
|
c = cLR
|
|
case hLRN:
|
|
c = cLRN
|
|
case hLRND:
|
|
c = cLRND
|
|
default:
|
|
return 0, errHeaderByte
|
|
}
|
|
return
|
|
}
|
|
|
|
// uncompressedHeaderLen provides the length of an uncompressed header
|
|
const uncompressedHeaderLen = 3
|
|
|
|
// headerLen returns the length of the LZMA2 header for a given chunk
|
|
// type.
|
|
func headerLen(c chunkType) int {
|
|
switch c {
|
|
case cEOS:
|
|
return 1
|
|
case cU, cUD:
|
|
return uncompressedHeaderLen
|
|
case cL, cLR:
|
|
return 5
|
|
case cLRN, cLRND:
|
|
return 6
|
|
}
|
|
panic(fmt.Errorf("unsupported chunk type %d", c))
|
|
}
|
|
|
|
// chunkHeader represents the contents of a chunk header.
|
|
type chunkHeader struct {
|
|
ctype chunkType
|
|
uncompressed uint32
|
|
compressed uint16
|
|
props Properties
|
|
}
|
|
|
|
// String returns a string representation of the chunk header.
|
|
func (h *chunkHeader) String() string {
|
|
return fmt.Sprintf("%s %d %d %s", h.ctype, h.uncompressed,
|
|
h.compressed, &h.props)
|
|
}
|
|
|
|
// UnmarshalBinary reads the content of the chunk header from the data
|
|
// slice. The slice must have the correct length.
|
|
func (h *chunkHeader) UnmarshalBinary(data []byte) error {
|
|
if len(data) == 0 {
|
|
return errors.New("no data")
|
|
}
|
|
c, err := headerChunkType(data[0])
|
|
if err != nil {
|
|
return err
|
|
}
|
|
|
|
n := headerLen(c)
|
|
if len(data) < n {
|
|
return errors.New("incomplete data")
|
|
}
|
|
if len(data) > n {
|
|
return errors.New("invalid data length")
|
|
}
|
|
|
|
*h = chunkHeader{ctype: c}
|
|
if c == cEOS {
|
|
return nil
|
|
}
|
|
|
|
h.uncompressed = uint32(uint16BE(data[1:3]))
|
|
if c <= cU {
|
|
return nil
|
|
}
|
|
h.uncompressed |= uint32(data[0]&^hLRND) << 16
|
|
|
|
h.compressed = uint16BE(data[3:5])
|
|
if c <= cLR {
|
|
return nil
|
|
}
|
|
|
|
h.props, err = PropertiesForCode(data[5])
|
|
return err
|
|
}
|
|
|
|
// MarshalBinary encodes the chunk header value. The function checks
|
|
// whether the content of the chunk header is correct.
|
|
func (h *chunkHeader) MarshalBinary() (data []byte, err error) {
|
|
if h.ctype > cLRND {
|
|
return nil, errors.New("invalid chunk type")
|
|
}
|
|
if err = h.props.verify(); err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
data = make([]byte, headerLen(h.ctype))
|
|
|
|
switch h.ctype {
|
|
case cEOS:
|
|
return data, nil
|
|
case cUD:
|
|
data[0] = hUD
|
|
case cU:
|
|
data[0] = hU
|
|
case cL:
|
|
data[0] = hL
|
|
case cLR:
|
|
data[0] = hLR
|
|
case cLRN:
|
|
data[0] = hLRN
|
|
case cLRND:
|
|
data[0] = hLRND
|
|
}
|
|
|
|
putUint16BE(data[1:3], uint16(h.uncompressed))
|
|
if h.ctype <= cU {
|
|
return data, nil
|
|
}
|
|
data[0] |= byte(h.uncompressed>>16) &^ hLRND
|
|
|
|
putUint16BE(data[3:5], h.compressed)
|
|
if h.ctype <= cLR {
|
|
return data, nil
|
|
}
|
|
|
|
data[5] = h.props.Code()
|
|
return data, nil
|
|
}
|
|
|
|
// readChunkHeader reads the chunk header from the IO reader.
|
|
func readChunkHeader(r io.Reader) (h *chunkHeader, err error) {
|
|
p := make([]byte, 1, 6)
|
|
if _, err = io.ReadFull(r, p); err != nil {
|
|
return
|
|
}
|
|
c, err := headerChunkType(p[0])
|
|
if err != nil {
|
|
return
|
|
}
|
|
p = p[:headerLen(c)]
|
|
if _, err = io.ReadFull(r, p[1:]); err != nil {
|
|
return
|
|
}
|
|
h = new(chunkHeader)
|
|
if err = h.UnmarshalBinary(p); err != nil {
|
|
return nil, err
|
|
}
|
|
return h, nil
|
|
}
|
|
|
|
// uint16BE converts a big-endian uint16 representation to an uint16
|
|
// value.
|
|
func uint16BE(p []byte) uint16 {
|
|
return uint16(p[0])<<8 | uint16(p[1])
|
|
}
|
|
|
|
// putUint16BE puts the big-endian uint16 presentation into the given
|
|
// slice.
|
|
func putUint16BE(p []byte, x uint16) {
|
|
p[0] = byte(x >> 8)
|
|
p[1] = byte(x)
|
|
}
|
|
|
|
// chunkState is used to manage the state of the chunks
|
|
type chunkState byte
|
|
|
|
// start and stop define the initial and terminating state of the chunk
|
|
// state
|
|
const (
|
|
start chunkState = 'S'
|
|
stop = 'T'
|
|
)
|
|
|
|
// errors for the chunk state handling
|
|
var (
|
|
errChunkType = errors.New("lzma: unexpected chunk type")
|
|
errState = errors.New("lzma: wrong chunk state")
|
|
)
|
|
|
|
// next transitions state based on chunk type input
|
|
func (c *chunkState) next(ctype chunkType) error {
|
|
switch *c {
|
|
// start state
|
|
case 'S':
|
|
switch ctype {
|
|
case cEOS:
|
|
*c = 'T'
|
|
case cUD:
|
|
*c = 'R'
|
|
case cLRND:
|
|
*c = 'L'
|
|
default:
|
|
return errChunkType
|
|
}
|
|
// normal LZMA mode
|
|
case 'L':
|
|
switch ctype {
|
|
case cEOS:
|
|
*c = 'T'
|
|
case cUD:
|
|
*c = 'R'
|
|
case cU:
|
|
*c = 'U'
|
|
case cL, cLR, cLRN, cLRND:
|
|
break
|
|
default:
|
|
return errChunkType
|
|
}
|
|
// reset required
|
|
case 'R':
|
|
switch ctype {
|
|
case cEOS:
|
|
*c = 'T'
|
|
case cUD, cU:
|
|
break
|
|
case cLRN, cLRND:
|
|
*c = 'L'
|
|
default:
|
|
return errChunkType
|
|
}
|
|
// uncompressed
|
|
case 'U':
|
|
switch ctype {
|
|
case cEOS:
|
|
*c = 'T'
|
|
case cUD:
|
|
*c = 'R'
|
|
case cU:
|
|
break
|
|
case cL, cLR, cLRN, cLRND:
|
|
*c = 'L'
|
|
default:
|
|
return errChunkType
|
|
}
|
|
// terminal state
|
|
case 'T':
|
|
return errChunkType
|
|
default:
|
|
return errState
|
|
}
|
|
return nil
|
|
}
|
|
|
|
// defaultChunkType returns the default chunk type for each chunk state.
|
|
func (c chunkState) defaultChunkType() chunkType {
|
|
switch c {
|
|
case 'S':
|
|
return cLRND
|
|
case 'L', 'U':
|
|
return cL
|
|
case 'R':
|
|
return cLRN
|
|
default:
|
|
// no error
|
|
return cEOS
|
|
}
|
|
}
|
|
|
|
// maxDictCap defines the maximum dictionary capacity supported by the
|
|
// LZMA2 dictionary capacity encoding.
|
|
const maxDictCap = 1<<32 - 1
|
|
|
|
// maxDictCapCode defines the maximum dictionary capacity code.
|
|
const maxDictCapCode = 40
|
|
|
|
// The function decodes the dictionary capacity byte, but doesn't change
|
|
// for the correct range of the given byte.
|
|
func decodeDictCap(c byte) int64 {
|
|
return (2 | int64(c)&1) << (11 + (c>>1)&0x1f)
|
|
}
|
|
|
|
// DecodeDictCap decodes the encoded dictionary capacity. The function
|
|
// returns an error if the code is out of range.
|
|
func DecodeDictCap(c byte) (n int64, err error) {
|
|
if c >= maxDictCapCode {
|
|
if c == maxDictCapCode {
|
|
return maxDictCap, nil
|
|
}
|
|
return 0, errors.New("lzma: invalid dictionary size code")
|
|
}
|
|
return decodeDictCap(c), nil
|
|
}
|
|
|
|
// EncodeDictCap encodes a dictionary capacity. The function returns the
|
|
// code for the capacity that is greater or equal n. If n exceeds the
|
|
// maximum support dictionary capacity, the maximum value is returned.
|
|
func EncodeDictCap(n int64) byte {
|
|
a, b := byte(0), byte(40)
|
|
for a < b {
|
|
c := a + (b-a)>>1
|
|
m := decodeDictCap(c)
|
|
if n <= m {
|
|
if n == m {
|
|
return c
|
|
}
|
|
b = c
|
|
} else {
|
|
a = c + 1
|
|
}
|
|
}
|
|
return a
|
|
}
|