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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383
|
// Copyright 2021 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package pkgbits
import (
"bytes"
"crypto/md5"
"encoding/binary"
"go/constant"
"io"
"math/big"
"runtime"
)
// currentVersion is the current version number.
//
// - v0: initial prototype
//
// - v1: adds the flags uint32 word
const currentVersion uint32 = 1
// A PkgEncoder provides methods for encoding a package's Unified IR
// export data.
type PkgEncoder struct {
// elems holds the bitstream for previously encoded elements.
elems [numRelocs][]string
// stringsIdx maps previously encoded strings to their index within
// the RelocString section, to allow deduplication. That is,
// elems[RelocString][stringsIdx[s]] == s (if present).
stringsIdx map[string]Index
// syncFrames is the number of frames to write at each sync
// marker. A negative value means sync markers are omitted.
syncFrames int
}
// SyncMarkers reports whether pw uses sync markers.
func (pw *PkgEncoder) SyncMarkers() bool { return pw.syncFrames >= 0 }
// NewPkgEncoder returns an initialized PkgEncoder.
//
// syncFrames is the number of caller frames that should be serialized
// at Sync points. Serializing additional frames results in larger
// export data files, but can help diagnosing desync errors in
// higher-level Unified IR reader/writer code. If syncFrames is
// negative, then sync markers are omitted entirely.
func NewPkgEncoder(syncFrames int) PkgEncoder {
return PkgEncoder{
stringsIdx: make(map[string]Index),
syncFrames: syncFrames,
}
}
// DumpTo writes the package's encoded data to out0 and returns the
// package fingerprint.
func (pw *PkgEncoder) DumpTo(out0 io.Writer) (fingerprint [8]byte) {
h := md5.New()
out := io.MultiWriter(out0, h)
writeUint32 := func(x uint32) {
assert(binary.Write(out, binary.LittleEndian, x) == nil)
}
writeUint32(currentVersion)
var flags uint32
if pw.SyncMarkers() {
flags |= flagSyncMarkers
}
writeUint32(flags)
// Write elemEndsEnds.
var sum uint32
for _, elems := range &pw.elems {
sum += uint32(len(elems))
writeUint32(sum)
}
// Write elemEnds.
sum = 0
for _, elems := range &pw.elems {
for _, elem := range elems {
sum += uint32(len(elem))
writeUint32(sum)
}
}
// Write elemData.
for _, elems := range &pw.elems {
for _, elem := range elems {
_, err := io.WriteString(out, elem)
assert(err == nil)
}
}
// Write fingerprint.
copy(fingerprint[:], h.Sum(nil))
_, err := out0.Write(fingerprint[:])
assert(err == nil)
return
}
// StringIdx adds a string value to the strings section, if not
// already present, and returns its index.
func (pw *PkgEncoder) StringIdx(s string) Index {
if idx, ok := pw.stringsIdx[s]; ok {
assert(pw.elems[RelocString][idx] == s)
return idx
}
idx := Index(len(pw.elems[RelocString]))
pw.elems[RelocString] = append(pw.elems[RelocString], s)
pw.stringsIdx[s] = idx
return idx
}
// NewEncoder returns an Encoder for a new element within the given
// section, and encodes the given SyncMarker as the start of the
// element bitstream.
func (pw *PkgEncoder) NewEncoder(k RelocKind, marker SyncMarker) Encoder {
e := pw.NewEncoderRaw(k)
e.Sync(marker)
return e
}
// NewEncoderRaw returns an Encoder for a new element within the given
// section.
//
// Most callers should use NewEncoder instead.
func (pw *PkgEncoder) NewEncoderRaw(k RelocKind) Encoder {
idx := Index(len(pw.elems[k]))
pw.elems[k] = append(pw.elems[k], "") // placeholder
return Encoder{
p: pw,
k: k,
Idx: idx,
}
}
// An Encoder provides methods for encoding an individual element's
// bitstream data.
type Encoder struct {
p *PkgEncoder
Relocs []RelocEnt
RelocMap map[RelocEnt]uint32
Data bytes.Buffer // accumulated element bitstream data
encodingRelocHeader bool
k RelocKind
Idx Index // index within relocation section
}
// Flush finalizes the element's bitstream and returns its Index.
func (w *Encoder) Flush() Index {
var sb bytes.Buffer // TODO(mdempsky): strings.Builder after #44505 is resolved
// Backup the data so we write the relocations at the front.
var tmp bytes.Buffer
io.Copy(&tmp, &w.Data)
// TODO(mdempsky): Consider writing these out separately so they're
// easier to strip, along with function bodies, so that we can prune
// down to just the data that's relevant to go/types.
if w.encodingRelocHeader {
panic("encodingRelocHeader already true; recursive flush?")
}
w.encodingRelocHeader = true
w.Sync(SyncRelocs)
w.Len(len(w.Relocs))
for _, rEnt := range w.Relocs {
w.Sync(SyncReloc)
w.Len(int(rEnt.Kind))
w.Len(int(rEnt.Idx))
}
io.Copy(&sb, &w.Data)
io.Copy(&sb, &tmp)
w.p.elems[w.k][w.Idx] = sb.String()
return w.Idx
}
func (w *Encoder) checkErr(err error) {
if err != nil {
errorf("unexpected encoding error: %v", err)
}
}
func (w *Encoder) rawUvarint(x uint64) {
var buf [binary.MaxVarintLen64]byte
n := binary.PutUvarint(buf[:], x)
_, err := w.Data.Write(buf[:n])
w.checkErr(err)
}
func (w *Encoder) rawVarint(x int64) {
// Zig-zag encode.
ux := uint64(x) << 1
if x < 0 {
ux = ^ux
}
w.rawUvarint(ux)
}
func (w *Encoder) rawReloc(r RelocKind, idx Index) int {
e := RelocEnt{r, idx}
if w.RelocMap != nil {
if i, ok := w.RelocMap[e]; ok {
return int(i)
}
} else {
w.RelocMap = make(map[RelocEnt]uint32)
}
i := len(w.Relocs)
w.RelocMap[e] = uint32(i)
w.Relocs = append(w.Relocs, e)
return i
}
func (w *Encoder) Sync(m SyncMarker) {
if !w.p.SyncMarkers() {
return
}
// Writing out stack frame string references requires working
// relocations, but writing out the relocations themselves involves
// sync markers. To prevent infinite recursion, we simply trim the
// stack frame for sync markers within the relocation header.
var frames []string
if !w.encodingRelocHeader && w.p.syncFrames > 0 {
pcs := make([]uintptr, w.p.syncFrames)
n := runtime.Callers(2, pcs)
frames = fmtFrames(pcs[:n]...)
}
// TODO(mdempsky): Save space by writing out stack frames as a
// linked list so we can share common stack frames.
w.rawUvarint(uint64(m))
w.rawUvarint(uint64(len(frames)))
for _, frame := range frames {
w.rawUvarint(uint64(w.rawReloc(RelocString, w.p.StringIdx(frame))))
}
}
// Bool encodes and writes a bool value into the element bitstream,
// and then returns the bool value.
//
// For simple, 2-alternative encodings, the idiomatic way to call Bool
// is something like:
//
// if w.Bool(x != 0) {
// // alternative #1
// } else {
// // alternative #2
// }
//
// For multi-alternative encodings, use Code instead.
func (w *Encoder) Bool(b bool) bool {
w.Sync(SyncBool)
var x byte
if b {
x = 1
}
err := w.Data.WriteByte(x)
w.checkErr(err)
return b
}
// Int64 encodes and writes an int64 value into the element bitstream.
func (w *Encoder) Int64(x int64) {
w.Sync(SyncInt64)
w.rawVarint(x)
}
// Uint64 encodes and writes a uint64 value into the element bitstream.
func (w *Encoder) Uint64(x uint64) {
w.Sync(SyncUint64)
w.rawUvarint(x)
}
// Len encodes and writes a non-negative int value into the element bitstream.
func (w *Encoder) Len(x int) { assert(x >= 0); w.Uint64(uint64(x)) }
// Int encodes and writes an int value into the element bitstream.
func (w *Encoder) Int(x int) { w.Int64(int64(x)) }
// Len encodes and writes a uint value into the element bitstream.
func (w *Encoder) Uint(x uint) { w.Uint64(uint64(x)) }
// Reloc encodes and writes a relocation for the given (section,
// index) pair into the element bitstream.
//
// Note: Only the index is formally written into the element
// bitstream, so bitstream decoders must know from context which
// section an encoded relocation refers to.
func (w *Encoder) Reloc(r RelocKind, idx Index) {
w.Sync(SyncUseReloc)
w.Len(w.rawReloc(r, idx))
}
// Code encodes and writes a Code value into the element bitstream.
func (w *Encoder) Code(c Code) {
w.Sync(c.Marker())
w.Len(c.Value())
}
// String encodes and writes a string value into the element
// bitstream.
//
// Internally, strings are deduplicated by adding them to the strings
// section (if not already present), and then writing a relocation
// into the element bitstream.
func (w *Encoder) String(s string) {
w.Sync(SyncString)
w.Reloc(RelocString, w.p.StringIdx(s))
}
// Strings encodes and writes a variable-length slice of strings into
// the element bitstream.
func (w *Encoder) Strings(ss []string) {
w.Len(len(ss))
for _, s := range ss {
w.String(s)
}
}
// Value encodes and writes a constant.Value into the element
// bitstream.
func (w *Encoder) Value(val constant.Value) {
w.Sync(SyncValue)
if w.Bool(val.Kind() == constant.Complex) {
w.scalar(constant.Real(val))
w.scalar(constant.Imag(val))
} else {
w.scalar(val)
}
}
func (w *Encoder) scalar(val constant.Value) {
switch v := constant.Val(val).(type) {
default:
errorf("unhandled %v (%v)", val, val.Kind())
case bool:
w.Code(ValBool)
w.Bool(v)
case string:
w.Code(ValString)
w.String(v)
case int64:
w.Code(ValInt64)
w.Int64(v)
case *big.Int:
w.Code(ValBigInt)
w.bigInt(v)
case *big.Rat:
w.Code(ValBigRat)
w.bigInt(v.Num())
w.bigInt(v.Denom())
case *big.Float:
w.Code(ValBigFloat)
w.bigFloat(v)
}
}
func (w *Encoder) bigInt(v *big.Int) {
b := v.Bytes()
w.String(string(b)) // TODO: More efficient encoding.
w.Bool(v.Sign() < 0)
}
func (w *Encoder) bigFloat(v *big.Float) {
b := v.Append(nil, 'p', -1)
w.String(string(b)) // TODO: More efficient encoding.
}
|