1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
|
package ingest
import (
"encoding/binary"
"fmt"
"hash/crc32"
"io"
"os"
"codeberg.org/lindenii/furgit/internal/compress/zlib"
"codeberg.org/lindenii/furgit/objectid"
"codeberg.org/lindenii/furgit/objecttype"
)
// maybeFixThin appends missing bases and rewrites pack header/trailer when needed.
func maybeFixThin(state *ingestState) error {
if len(state.unresolvedRefDeltas) == 0 {
return nil
}
if !state.fixThin {
return &ErrThinPackUnresolved{Count: len(state.unresolvedRefDeltas)}
}
if state.base == nil {
return &ErrThinPackUnresolved{Count: len(state.unresolvedRefDeltas)}
}
hashSize := int64(state.algo.Size())
info, err := state.packFile.Stat()
if err != nil {
return err
}
size := info.Size()
if size < hashSize {
return fmt.Errorf("format/pack/ingest: pack too short to trim trailer")
}
newEnd := size - hashSize
if err := state.packFile.Truncate(newEnd); err != nil {
return err
}
state.stream.consumed = uint64(newEnd)
baseIDs := unresolvedThinBaseIDs(state)
for _, id := range baseIDs {
ty, content, err := state.base.ReadBytesContent(id)
if err != nil {
continue
}
if _, err := appendBaseObject(state, id, ty, content); err != nil {
return err
}
state.thinFixed = true
}
if err := rewritePackHeaderAndTrailer(state); err != nil {
return err
}
return nil
}
// appendBaseObject appends one base object as a new packed non-delta entry.
func appendBaseObject(state *ingestState, id objectid.ObjectID, realType objecttype.Type, content []byte) (int, error) {
start := state.stream.consumed
header := encodePackEntryHeader(realType, int64(len(content)))
if _, err := state.packFile.WriteAt(header, int64(start)); err != nil {
return 0, err
}
section := &fileSectionWriter{file: state.packFile, off: int64(start) + int64(len(header))}
crc := crc32.NewIEEE()
_, _ = crc.Write(header)
counting := &countingWriter{dst: section}
zw := zlib.NewWriter(io.MultiWriter(counting, crc))
if _, err := zw.Write(content); err != nil {
return 0, err
}
if err := zw.Close(); err != nil {
return 0, err
}
packedLen := uint64(len(header)) + uint64(counting.n)
end := start + packedLen
state.stream.consumed = end
record := objectRecord{
offset: start,
headerLen: uint32(len(header)),
packedLen: packedLen,
crc32: crc.Sum32(),
packedType: realType,
realType: realType,
declaredSize: int64(len(content)),
dataOffset: start + uint64(len(header)),
objectID: id,
resolved: true,
}
recordIdx := len(state.records)
state.records = append(state.records, record)
state.offsetToRecord[start] = recordIdx
state.objectToRecord[id] = recordIdx
state.baseCache.add(recordIdx, realType, content)
return recordIdx, nil
}
// fileSectionWriter writes sequentially to file via WriteAt at one base offset.
type fileSectionWriter struct {
file *os.File
off int64
pos int64
}
// Write writes src at current section position.
func (writer *fileSectionWriter) Write(src []byte) (int, error) {
if len(src) == 0 {
return 0, nil
}
n, err := writer.file.WriteAt(src, writer.off+writer.pos)
writer.pos += int64(n)
return n, err
}
// countingWriter counts bytes written to dst.
type countingWriter struct {
dst io.Writer
n int
}
// Write writes src to dst and tracks output byte count.
func (writer *countingWriter) Write(src []byte) (int, error) {
n, err := writer.dst.Write(src)
writer.n += n
return n, err
}
// rewritePackHeaderAndTrailer rewrites object count and trailer hash using ReadAt/WriteAt.
func rewritePackHeaderAndTrailer(state *ingestState) error {
var countRaw [4]byte
binary.BigEndian.PutUint32(countRaw[:], uint32(len(state.records)))
if _, err := state.packFile.WriteAt(countRaw[:], 8); err != nil {
return err
}
info, err := state.packFile.Stat()
if err != nil {
return err
}
endWithoutTrailer := info.Size()
hashImpl, err := state.algo.New()
if err != nil {
return err
}
var (
buf [128 << 10]byte
pos int64
)
for pos < endWithoutTrailer {
want := int64(len(buf))
remaining := endWithoutTrailer - pos
if remaining < want {
want = remaining
}
n, err := state.packFile.ReadAt(buf[:want], pos)
if err != nil && err != io.EOF {
return err
}
if n == 0 {
return io.ErrUnexpectedEOF
}
_, _ = hashImpl.Write(buf[:n])
pos += int64(n)
}
sum := hashImpl.Sum(nil)
if _, err := state.packFile.WriteAt(sum, endWithoutTrailer); err != nil {
return err
}
packHash, err := objectid.FromBytes(state.algo, sum)
if err != nil {
return err
}
state.packHash = packHash
state.objectCountHeader = uint32(len(state.records))
state.stream.consumed = uint64(endWithoutTrailer + int64(len(sum)))
return nil
}
// encodePackEntryHeader encodes one non-delta packed entry header.
func encodePackEntryHeader(ty objecttype.Type, size int64) []byte {
var out [16]byte
n := 0
s := uint64(size)
c := byte((uint8(ty) << 4) | byte(s&0x0f))
s >>= 4
for s != 0 {
out[n] = c | 0x80
n++
c = byte(s & 0x7f)
s >>= 7
}
out[n] = c
n++
return append([]byte(nil), out[:n]...)
}
|
