summaryrefslogtreecommitdiff
path: root/libgo/go/encoding/gob/decode.go
diff options
context:
space:
mode:
Diffstat (limited to 'libgo/go/encoding/gob/decode.go')
-rw-r--r--libgo/go/encoding/gob/decode.go1279
1 files changed, 1279 insertions, 0 deletions
diff --git a/libgo/go/encoding/gob/decode.go b/libgo/go/encoding/gob/decode.go
new file mode 100644
index 00000000000..1515d1286d0
--- /dev/null
+++ b/libgo/go/encoding/gob/decode.go
@@ -0,0 +1,1279 @@
+// Copyright 2009 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 gob
+
+// TODO(rsc): When garbage collector changes, revisit
+// the allocations in this file that use unsafe.Pointer.
+
+import (
+ "bytes"
+ "errors"
+ "io"
+ "math"
+ "reflect"
+ "unsafe"
+)
+
+var (
+ errBadUint = errors.New("gob: encoded unsigned integer out of range")
+ errBadType = errors.New("gob: unknown type id or corrupted data")
+ errRange = errors.New("gob: bad data: field numbers out of bounds")
+)
+
+// decoderState is the execution state of an instance of the decoder. A new state
+// is created for nested objects.
+type decoderState struct {
+ dec *Decoder
+ // The buffer is stored with an extra indirection because it may be replaced
+ // if we load a type during decode (when reading an interface value).
+ b *bytes.Buffer
+ fieldnum int // the last field number read.
+ buf []byte
+ next *decoderState // for free list
+}
+
+// We pass the bytes.Buffer separately for easier testing of the infrastructure
+// without requiring a full Decoder.
+func (dec *Decoder) newDecoderState(buf *bytes.Buffer) *decoderState {
+ d := dec.freeList
+ if d == nil {
+ d = new(decoderState)
+ d.dec = dec
+ d.buf = make([]byte, uint64Size)
+ } else {
+ dec.freeList = d.next
+ }
+ d.b = buf
+ return d
+}
+
+func (dec *Decoder) freeDecoderState(d *decoderState) {
+ d.next = dec.freeList
+ dec.freeList = d
+}
+
+func overflow(name string) error {
+ return errors.New(`value for "` + name + `" out of range`)
+}
+
+// decodeUintReader reads an encoded unsigned integer from an io.Reader.
+// Used only by the Decoder to read the message length.
+func decodeUintReader(r io.Reader, buf []byte) (x uint64, width int, err error) {
+ width = 1
+ _, err = r.Read(buf[0:width])
+ if err != nil {
+ return
+ }
+ b := buf[0]
+ if b <= 0x7f {
+ return uint64(b), width, nil
+ }
+ n := -int(int8(b))
+ if n > uint64Size {
+ err = errBadUint
+ return
+ }
+ width, err = io.ReadFull(r, buf[0:n])
+ if err != nil {
+ if err == io.EOF {
+ err = io.ErrUnexpectedEOF
+ }
+ return
+ }
+ // Could check that the high byte is zero but it's not worth it.
+ for _, b := range buf[0:width] {
+ x = x<<8 | uint64(b)
+ }
+ width++ // +1 for length byte
+ return
+}
+
+// decodeUint reads an encoded unsigned integer from state.r.
+// Does not check for overflow.
+func (state *decoderState) decodeUint() (x uint64) {
+ b, err := state.b.ReadByte()
+ if err != nil {
+ error_(err)
+ }
+ if b <= 0x7f {
+ return uint64(b)
+ }
+ n := -int(int8(b))
+ if n > uint64Size {
+ error_(errBadUint)
+ }
+ width, err := state.b.Read(state.buf[0:n])
+ if err != nil {
+ error_(err)
+ }
+ // Don't need to check error; it's safe to loop regardless.
+ // Could check that the high byte is zero but it's not worth it.
+ for _, b := range state.buf[0:width] {
+ x = x<<8 | uint64(b)
+ }
+ return x
+}
+
+// decodeInt reads an encoded signed integer from state.r.
+// Does not check for overflow.
+func (state *decoderState) decodeInt() int64 {
+ x := state.decodeUint()
+ if x&1 != 0 {
+ return ^int64(x >> 1)
+ }
+ return int64(x >> 1)
+}
+
+// decOp is the signature of a decoding operator for a given type.
+type decOp func(i *decInstr, state *decoderState, p unsafe.Pointer)
+
+// The 'instructions' of the decoding machine
+type decInstr struct {
+ op decOp
+ field int // field number of the wire type
+ indir int // how many pointer indirections to reach the value in the struct
+ offset uintptr // offset in the structure of the field to encode
+ ovfl error // error message for overflow/underflow (for arrays, of the elements)
+}
+
+// Since the encoder writes no zeros, if we arrive at a decoder we have
+// a value to extract and store. The field number has already been read
+// (it's how we knew to call this decoder).
+// Each decoder is responsible for handling any indirections associated
+// with the data structure. If any pointer so reached is nil, allocation must
+// be done.
+
+// Walk the pointer hierarchy, allocating if we find a nil. Stop one before the end.
+func decIndirect(p unsafe.Pointer, indir int) unsafe.Pointer {
+ for ; indir > 1; indir-- {
+ if *(*unsafe.Pointer)(p) == nil {
+ // Allocation required
+ *(*unsafe.Pointer)(p) = unsafe.Pointer(new(unsafe.Pointer))
+ }
+ p = *(*unsafe.Pointer)(p)
+ }
+ return p
+}
+
+// ignoreUint discards a uint value with no destination.
+func ignoreUint(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ state.decodeUint()
+}
+
+// ignoreTwoUints discards a uint value with no destination. It's used to skip
+// complex values.
+func ignoreTwoUints(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ state.decodeUint()
+ state.decodeUint()
+}
+
+// decBool decodes a uint and stores it as a boolean through p.
+func decBool(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ if i.indir > 0 {
+ if *(*unsafe.Pointer)(p) == nil {
+ *(*unsafe.Pointer)(p) = unsafe.Pointer(new(bool))
+ }
+ p = *(*unsafe.Pointer)(p)
+ }
+ *(*bool)(p) = state.decodeUint() != 0
+}
+
+// decInt8 decodes an integer and stores it as an int8 through p.
+func decInt8(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ if i.indir > 0 {
+ if *(*unsafe.Pointer)(p) == nil {
+ *(*unsafe.Pointer)(p) = unsafe.Pointer(new(int8))
+ }
+ p = *(*unsafe.Pointer)(p)
+ }
+ v := state.decodeInt()
+ if v < math.MinInt8 || math.MaxInt8 < v {
+ error_(i.ovfl)
+ } else {
+ *(*int8)(p) = int8(v)
+ }
+}
+
+// decUint8 decodes an unsigned integer and stores it as a uint8 through p.
+func decUint8(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ if i.indir > 0 {
+ if *(*unsafe.Pointer)(p) == nil {
+ *(*unsafe.Pointer)(p) = unsafe.Pointer(new(uint8))
+ }
+ p = *(*unsafe.Pointer)(p)
+ }
+ v := state.decodeUint()
+ if math.MaxUint8 < v {
+ error_(i.ovfl)
+ } else {
+ *(*uint8)(p) = uint8(v)
+ }
+}
+
+// decInt16 decodes an integer and stores it as an int16 through p.
+func decInt16(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ if i.indir > 0 {
+ if *(*unsafe.Pointer)(p) == nil {
+ *(*unsafe.Pointer)(p) = unsafe.Pointer(new(int16))
+ }
+ p = *(*unsafe.Pointer)(p)
+ }
+ v := state.decodeInt()
+ if v < math.MinInt16 || math.MaxInt16 < v {
+ error_(i.ovfl)
+ } else {
+ *(*int16)(p) = int16(v)
+ }
+}
+
+// decUint16 decodes an unsigned integer and stores it as a uint16 through p.
+func decUint16(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ if i.indir > 0 {
+ if *(*unsafe.Pointer)(p) == nil {
+ *(*unsafe.Pointer)(p) = unsafe.Pointer(new(uint16))
+ }
+ p = *(*unsafe.Pointer)(p)
+ }
+ v := state.decodeUint()
+ if math.MaxUint16 < v {
+ error_(i.ovfl)
+ } else {
+ *(*uint16)(p) = uint16(v)
+ }
+}
+
+// decInt32 decodes an integer and stores it as an int32 through p.
+func decInt32(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ if i.indir > 0 {
+ if *(*unsafe.Pointer)(p) == nil {
+ *(*unsafe.Pointer)(p) = unsafe.Pointer(new(int32))
+ }
+ p = *(*unsafe.Pointer)(p)
+ }
+ v := state.decodeInt()
+ if v < math.MinInt32 || math.MaxInt32 < v {
+ error_(i.ovfl)
+ } else {
+ *(*int32)(p) = int32(v)
+ }
+}
+
+// decUint32 decodes an unsigned integer and stores it as a uint32 through p.
+func decUint32(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ if i.indir > 0 {
+ if *(*unsafe.Pointer)(p) == nil {
+ *(*unsafe.Pointer)(p) = unsafe.Pointer(new(uint32))
+ }
+ p = *(*unsafe.Pointer)(p)
+ }
+ v := state.decodeUint()
+ if math.MaxUint32 < v {
+ error_(i.ovfl)
+ } else {
+ *(*uint32)(p) = uint32(v)
+ }
+}
+
+// decInt64 decodes an integer and stores it as an int64 through p.
+func decInt64(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ if i.indir > 0 {
+ if *(*unsafe.Pointer)(p) == nil {
+ *(*unsafe.Pointer)(p) = unsafe.Pointer(new(int64))
+ }
+ p = *(*unsafe.Pointer)(p)
+ }
+ *(*int64)(p) = int64(state.decodeInt())
+}
+
+// decUint64 decodes an unsigned integer and stores it as a uint64 through p.
+func decUint64(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ if i.indir > 0 {
+ if *(*unsafe.Pointer)(p) == nil {
+ *(*unsafe.Pointer)(p) = unsafe.Pointer(new(uint64))
+ }
+ p = *(*unsafe.Pointer)(p)
+ }
+ *(*uint64)(p) = uint64(state.decodeUint())
+}
+
+// Floating-point numbers are transmitted as uint64s holding the bits
+// of the underlying representation. They are sent byte-reversed, with
+// the exponent end coming out first, so integer floating point numbers
+// (for example) transmit more compactly. This routine does the
+// unswizzling.
+func floatFromBits(u uint64) float64 {
+ var v uint64
+ for i := 0; i < 8; i++ {
+ v <<= 8
+ v |= u & 0xFF
+ u >>= 8
+ }
+ return math.Float64frombits(v)
+}
+
+// storeFloat32 decodes an unsigned integer, treats it as a 32-bit floating-point
+// number, and stores it through p. It's a helper function for float32 and complex64.
+func storeFloat32(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ v := floatFromBits(state.decodeUint())
+ av := v
+ if av < 0 {
+ av = -av
+ }
+ // +Inf is OK in both 32- and 64-bit floats. Underflow is always OK.
+ if math.MaxFloat32 < av && av <= math.MaxFloat64 {
+ error_(i.ovfl)
+ } else {
+ *(*float32)(p) = float32(v)
+ }
+}
+
+// decFloat32 decodes an unsigned integer, treats it as a 32-bit floating-point
+// number, and stores it through p.
+func decFloat32(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ if i.indir > 0 {
+ if *(*unsafe.Pointer)(p) == nil {
+ *(*unsafe.Pointer)(p) = unsafe.Pointer(new(float32))
+ }
+ p = *(*unsafe.Pointer)(p)
+ }
+ storeFloat32(i, state, p)
+}
+
+// decFloat64 decodes an unsigned integer, treats it as a 64-bit floating-point
+// number, and stores it through p.
+func decFloat64(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ if i.indir > 0 {
+ if *(*unsafe.Pointer)(p) == nil {
+ *(*unsafe.Pointer)(p) = unsafe.Pointer(new(float64))
+ }
+ p = *(*unsafe.Pointer)(p)
+ }
+ *(*float64)(p) = floatFromBits(uint64(state.decodeUint()))
+}
+
+// decComplex64 decodes a pair of unsigned integers, treats them as a
+// pair of floating point numbers, and stores them as a complex64 through p.
+// The real part comes first.
+func decComplex64(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ if i.indir > 0 {
+ if *(*unsafe.Pointer)(p) == nil {
+ *(*unsafe.Pointer)(p) = unsafe.Pointer(new(complex64))
+ }
+ p = *(*unsafe.Pointer)(p)
+ }
+ storeFloat32(i, state, p)
+ storeFloat32(i, state, unsafe.Pointer(uintptr(p)+unsafe.Sizeof(float32(0))))
+}
+
+// decComplex128 decodes a pair of unsigned integers, treats them as a
+// pair of floating point numbers, and stores them as a complex128 through p.
+// The real part comes first.
+func decComplex128(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ if i.indir > 0 {
+ if *(*unsafe.Pointer)(p) == nil {
+ *(*unsafe.Pointer)(p) = unsafe.Pointer(new(complex128))
+ }
+ p = *(*unsafe.Pointer)(p)
+ }
+ real := floatFromBits(uint64(state.decodeUint()))
+ imag := floatFromBits(uint64(state.decodeUint()))
+ *(*complex128)(p) = complex(real, imag)
+}
+
+// decUint8Slice decodes a byte slice and stores through p a slice header
+// describing the data.
+// uint8 slices are encoded as an unsigned count followed by the raw bytes.
+func decUint8Slice(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ if i.indir > 0 {
+ if *(*unsafe.Pointer)(p) == nil {
+ *(*unsafe.Pointer)(p) = unsafe.Pointer(new([]uint8))
+ }
+ p = *(*unsafe.Pointer)(p)
+ }
+ n := int(state.decodeUint())
+ if n < 0 {
+ errorf("negative length decoding []byte")
+ }
+ slice := (*[]uint8)(p)
+ if cap(*slice) < n {
+ *slice = make([]uint8, n)
+ } else {
+ *slice = (*slice)[0:n]
+ }
+ if _, err := state.b.Read(*slice); err != nil {
+ errorf("error decoding []byte: %s", err)
+ }
+}
+
+// decString decodes byte array and stores through p a string header
+// describing the data.
+// Strings are encoded as an unsigned count followed by the raw bytes.
+func decString(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ if i.indir > 0 {
+ if *(*unsafe.Pointer)(p) == nil {
+ *(*unsafe.Pointer)(p) = unsafe.Pointer(new(string))
+ }
+ p = *(*unsafe.Pointer)(p)
+ }
+ b := make([]byte, state.decodeUint())
+ state.b.Read(b)
+ // It would be a shame to do the obvious thing here,
+ // *(*string)(p) = string(b)
+ // because we've already allocated the storage and this would
+ // allocate again and copy. So we do this ugly hack, which is even
+ // even more unsafe than it looks as it depends the memory
+ // representation of a string matching the beginning of the memory
+ // representation of a byte slice (a byte slice is longer).
+ *(*string)(p) = *(*string)(unsafe.Pointer(&b))
+}
+
+// ignoreUint8Array skips over the data for a byte slice value with no destination.
+func ignoreUint8Array(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ b := make([]byte, state.decodeUint())
+ state.b.Read(b)
+}
+
+// Execution engine
+
+// The encoder engine is an array of instructions indexed by field number of the incoming
+// decoder. It is executed with random access according to field number.
+type decEngine struct {
+ instr []decInstr
+ numInstr int // the number of active instructions
+}
+
+// allocate makes sure storage is available for an object of underlying type rtyp
+// that is indir levels of indirection through p.
+func allocate(rtyp reflect.Type, p uintptr, indir int) uintptr {
+ if indir == 0 {
+ return p
+ }
+ up := unsafe.Pointer(p)
+ if indir > 1 {
+ up = decIndirect(up, indir)
+ }
+ if *(*unsafe.Pointer)(up) == nil {
+ // Allocate object.
+ *(*unsafe.Pointer)(up) = unsafe.New(rtyp)
+ }
+ return *(*uintptr)(up)
+}
+
+// decodeSingle decodes a top-level value that is not a struct and stores it through p.
+// Such values are preceded by a zero, making them have the memory layout of a
+// struct field (although with an illegal field number).
+func (dec *Decoder) decodeSingle(engine *decEngine, ut *userTypeInfo, basep uintptr) (err error) {
+ state := dec.newDecoderState(&dec.buf)
+ state.fieldnum = singletonField
+ delta := int(state.decodeUint())
+ if delta != 0 {
+ errorf("decode: corrupted data: non-zero delta for singleton")
+ }
+ instr := &engine.instr[singletonField]
+ if instr.indir != ut.indir {
+ return errors.New("gob: internal error: inconsistent indirection")
+ }
+ ptr := unsafe.Pointer(basep) // offset will be zero
+ if instr.indir > 1 {
+ ptr = decIndirect(ptr, instr.indir)
+ }
+ instr.op(instr, state, ptr)
+ dec.freeDecoderState(state)
+ return nil
+}
+
+// decodeSingle decodes a top-level struct and stores it through p.
+// Indir is for the value, not the type. At the time of the call it may
+// differ from ut.indir, which was computed when the engine was built.
+// This state cannot arise for decodeSingle, which is called directly
+// from the user's value, not from the innards of an engine.
+func (dec *Decoder) decodeStruct(engine *decEngine, ut *userTypeInfo, p uintptr, indir int) {
+ p = allocate(ut.base, p, indir)
+ state := dec.newDecoderState(&dec.buf)
+ state.fieldnum = -1
+ basep := p
+ for state.b.Len() > 0 {
+ delta := int(state.decodeUint())
+ if delta < 0 {
+ errorf("decode: corrupted data: negative delta")
+ }
+ if delta == 0 { // struct terminator is zero delta fieldnum
+ break
+ }
+ fieldnum := state.fieldnum + delta
+ if fieldnum >= len(engine.instr) {
+ error_(errRange)
+ break
+ }
+ instr := &engine.instr[fieldnum]
+ p := unsafe.Pointer(basep + instr.offset)
+ if instr.indir > 1 {
+ p = decIndirect(p, instr.indir)
+ }
+ instr.op(instr, state, p)
+ state.fieldnum = fieldnum
+ }
+ dec.freeDecoderState(state)
+}
+
+// ignoreStruct discards the data for a struct with no destination.
+func (dec *Decoder) ignoreStruct(engine *decEngine) {
+ state := dec.newDecoderState(&dec.buf)
+ state.fieldnum = -1
+ for state.b.Len() > 0 {
+ delta := int(state.decodeUint())
+ if delta < 0 {
+ errorf("ignore decode: corrupted data: negative delta")
+ }
+ if delta == 0 { // struct terminator is zero delta fieldnum
+ break
+ }
+ fieldnum := state.fieldnum + delta
+ if fieldnum >= len(engine.instr) {
+ error_(errRange)
+ }
+ instr := &engine.instr[fieldnum]
+ instr.op(instr, state, unsafe.Pointer(nil))
+ state.fieldnum = fieldnum
+ }
+ dec.freeDecoderState(state)
+}
+
+// ignoreSingle discards the data for a top-level non-struct value with no
+// destination. It's used when calling Decode with a nil value.
+func (dec *Decoder) ignoreSingle(engine *decEngine) {
+ state := dec.newDecoderState(&dec.buf)
+ state.fieldnum = singletonField
+ delta := int(state.decodeUint())
+ if delta != 0 {
+ errorf("decode: corrupted data: non-zero delta for singleton")
+ }
+ instr := &engine.instr[singletonField]
+ instr.op(instr, state, unsafe.Pointer(nil))
+ dec.freeDecoderState(state)
+}
+
+// decodeArrayHelper does the work for decoding arrays and slices.
+func (dec *Decoder) decodeArrayHelper(state *decoderState, p uintptr, elemOp decOp, elemWid uintptr, length, elemIndir int, ovfl error) {
+ instr := &decInstr{elemOp, 0, elemIndir, 0, ovfl}
+ for i := 0; i < length; i++ {
+ up := unsafe.Pointer(p)
+ if elemIndir > 1 {
+ up = decIndirect(up, elemIndir)
+ }
+ elemOp(instr, state, up)
+ p += uintptr(elemWid)
+ }
+}
+
+// decodeArray decodes an array and stores it through p, that is, p points to the zeroth element.
+// The length is an unsigned integer preceding the elements. Even though the length is redundant
+// (it's part of the type), it's a useful check and is included in the encoding.
+func (dec *Decoder) decodeArray(atyp reflect.Type, state *decoderState, p uintptr, elemOp decOp, elemWid uintptr, length, indir, elemIndir int, ovfl error) {
+ if indir > 0 {
+ p = allocate(atyp, p, 1) // All but the last level has been allocated by dec.Indirect
+ }
+ if n := state.decodeUint(); n != uint64(length) {
+ errorf("length mismatch in decodeArray")
+ }
+ dec.decodeArrayHelper(state, p, elemOp, elemWid, length, elemIndir, ovfl)
+}
+
+// decodeIntoValue is a helper for map decoding. Since maps are decoded using reflection,
+// unlike the other items we can't use a pointer directly.
+func decodeIntoValue(state *decoderState, op decOp, indir int, v reflect.Value, ovfl error) reflect.Value {
+ instr := &decInstr{op, 0, indir, 0, ovfl}
+ up := unsafe.Pointer(unsafeAddr(v))
+ if indir > 1 {
+ up = decIndirect(up, indir)
+ }
+ op(instr, state, up)
+ return v
+}
+
+// decodeMap decodes a map and stores its header through p.
+// Maps are encoded as a length followed by key:value pairs.
+// Because the internals of maps are not visible to us, we must
+// use reflection rather than pointer magic.
+func (dec *Decoder) decodeMap(mtyp reflect.Type, state *decoderState, p uintptr, keyOp, elemOp decOp, indir, keyIndir, elemIndir int, ovfl error) {
+ if indir > 0 {
+ p = allocate(mtyp, p, 1) // All but the last level has been allocated by dec.Indirect
+ }
+ up := unsafe.Pointer(p)
+ if *(*unsafe.Pointer)(up) == nil { // maps are represented as a pointer in the runtime
+ // Allocate map.
+ *(*unsafe.Pointer)(up) = unsafe.Pointer(reflect.MakeMap(mtyp).Pointer())
+ }
+ // Maps cannot be accessed by moving addresses around the way
+ // that slices etc. can. We must recover a full reflection value for
+ // the iteration.
+ v := reflect.ValueOf(unsafe.Unreflect(mtyp, unsafe.Pointer(p)))
+ n := int(state.decodeUint())
+ for i := 0; i < n; i++ {
+ key := decodeIntoValue(state, keyOp, keyIndir, allocValue(mtyp.Key()), ovfl)
+ elem := decodeIntoValue(state, elemOp, elemIndir, allocValue(mtyp.Elem()), ovfl)
+ v.SetMapIndex(key, elem)
+ }
+}
+
+// ignoreArrayHelper does the work for discarding arrays and slices.
+func (dec *Decoder) ignoreArrayHelper(state *decoderState, elemOp decOp, length int) {
+ instr := &decInstr{elemOp, 0, 0, 0, errors.New("no error")}
+ for i := 0; i < length; i++ {
+ elemOp(instr, state, nil)
+ }
+}
+
+// ignoreArray discards the data for an array value with no destination.
+func (dec *Decoder) ignoreArray(state *decoderState, elemOp decOp, length int) {
+ if n := state.decodeUint(); n != uint64(length) {
+ errorf("length mismatch in ignoreArray")
+ }
+ dec.ignoreArrayHelper(state, elemOp, length)
+}
+
+// ignoreMap discards the data for a map value with no destination.
+func (dec *Decoder) ignoreMap(state *decoderState, keyOp, elemOp decOp) {
+ n := int(state.decodeUint())
+ keyInstr := &decInstr{keyOp, 0, 0, 0, errors.New("no error")}
+ elemInstr := &decInstr{elemOp, 0, 0, 0, errors.New("no error")}
+ for i := 0; i < n; i++ {
+ keyOp(keyInstr, state, nil)
+ elemOp(elemInstr, state, nil)
+ }
+}
+
+// decodeSlice decodes a slice and stores the slice header through p.
+// Slices are encoded as an unsigned length followed by the elements.
+func (dec *Decoder) decodeSlice(atyp reflect.Type, state *decoderState, p uintptr, elemOp decOp, elemWid uintptr, indir, elemIndir int, ovfl error) {
+ n := int(uintptr(state.decodeUint()))
+ if indir > 0 {
+ up := unsafe.Pointer(p)
+ if *(*unsafe.Pointer)(up) == nil {
+ // Allocate the slice header.
+ *(*unsafe.Pointer)(up) = unsafe.Pointer(new([]unsafe.Pointer))
+ }
+ p = *(*uintptr)(up)
+ }
+ // Allocate storage for the slice elements, that is, the underlying array,
+ // if the existing slice does not have the capacity.
+ // Always write a header at p.
+ hdrp := (*reflect.SliceHeader)(unsafe.Pointer(p))
+ if hdrp.Cap < n {
+ hdrp.Data = uintptr(unsafe.NewArray(atyp.Elem(), n))
+ hdrp.Cap = n
+ }
+ hdrp.Len = n
+ dec.decodeArrayHelper(state, hdrp.Data, elemOp, elemWid, n, elemIndir, ovfl)
+}
+
+// ignoreSlice skips over the data for a slice value with no destination.
+func (dec *Decoder) ignoreSlice(state *decoderState, elemOp decOp) {
+ dec.ignoreArrayHelper(state, elemOp, int(state.decodeUint()))
+}
+
+// setInterfaceValue sets an interface value to a concrete value,
+// but first it checks that the assignment will succeed.
+func setInterfaceValue(ivalue reflect.Value, value reflect.Value) {
+ if !value.Type().AssignableTo(ivalue.Type()) {
+ errorf("cannot assign value of type %s to %s", value.Type(), ivalue.Type())
+ }
+ ivalue.Set(value)
+}
+
+// decodeInterface decodes an interface value and stores it through p.
+// Interfaces are encoded as the name of a concrete type followed by a value.
+// If the name is empty, the value is nil and no value is sent.
+func (dec *Decoder) decodeInterface(ityp reflect.Type, state *decoderState, p uintptr, indir int) {
+ // Create a writable interface reflect.Value. We need one even for the nil case.
+ ivalue := allocValue(ityp)
+ // Read the name of the concrete type.
+ b := make([]byte, state.decodeUint())
+ state.b.Read(b)
+ name := string(b)
+ if name == "" {
+ // Copy the representation of the nil interface value to the target.
+ // This is horribly unsafe and special.
+ *(*[2]uintptr)(unsafe.Pointer(p)) = ivalue.InterfaceData()
+ return
+ }
+ // The concrete type must be registered.
+ typ, ok := nameToConcreteType[name]
+ if !ok {
+ errorf("name not registered for interface: %q", name)
+ }
+ // Read the type id of the concrete value.
+ concreteId := dec.decodeTypeSequence(true)
+ if concreteId < 0 {
+ error_(dec.err)
+ }
+ // Byte count of value is next; we don't care what it is (it's there
+ // in case we want to ignore the value by skipping it completely).
+ state.decodeUint()
+ // Read the concrete value.
+ value := allocValue(typ)
+ dec.decodeValue(concreteId, value)
+ if dec.err != nil {
+ error_(dec.err)
+ }
+ // Allocate the destination interface value.
+ if indir > 0 {
+ p = allocate(ityp, p, 1) // All but the last level has been allocated by dec.Indirect
+ }
+ // Assign the concrete value to the interface.
+ // Tread carefully; it might not satisfy the interface.
+ setInterfaceValue(ivalue, value)
+ // Copy the representation of the interface value to the target.
+ // This is horribly unsafe and special.
+ *(*[2]uintptr)(unsafe.Pointer(p)) = ivalue.InterfaceData()
+}
+
+// ignoreInterface discards the data for an interface value with no destination.
+func (dec *Decoder) ignoreInterface(state *decoderState) {
+ // Read the name of the concrete type.
+ b := make([]byte, state.decodeUint())
+ _, err := state.b.Read(b)
+ if err != nil {
+ error_(err)
+ }
+ id := dec.decodeTypeSequence(true)
+ if id < 0 {
+ error_(dec.err)
+ }
+ // At this point, the decoder buffer contains a delimited value. Just toss it.
+ state.b.Next(int(state.decodeUint()))
+}
+
+// decodeGobDecoder decodes something implementing the GobDecoder interface.
+// The data is encoded as a byte slice.
+func (dec *Decoder) decodeGobDecoder(state *decoderState, v reflect.Value) {
+ // Read the bytes for the value.
+ b := make([]byte, state.decodeUint())
+ _, err := state.b.Read(b)
+ if err != nil {
+ error_(err)
+ }
+ // We know it's a GobDecoder, so just call the method directly.
+ err = v.Interface().(GobDecoder).GobDecode(b)
+ if err != nil {
+ error_(err)
+ }
+}
+
+// ignoreGobDecoder discards the data for a GobDecoder value with no destination.
+func (dec *Decoder) ignoreGobDecoder(state *decoderState) {
+ // Read the bytes for the value.
+ b := make([]byte, state.decodeUint())
+ _, err := state.b.Read(b)
+ if err != nil {
+ error_(err)
+ }
+}
+
+// Index by Go types.
+var decOpTable = [...]decOp{
+ reflect.Bool: decBool,
+ reflect.Int8: decInt8,
+ reflect.Int16: decInt16,
+ reflect.Int32: decInt32,
+ reflect.Int64: decInt64,
+ reflect.Uint8: decUint8,
+ reflect.Uint16: decUint16,
+ reflect.Uint32: decUint32,
+ reflect.Uint64: decUint64,
+ reflect.Float32: decFloat32,
+ reflect.Float64: decFloat64,
+ reflect.Complex64: decComplex64,
+ reflect.Complex128: decComplex128,
+ reflect.String: decString,
+}
+
+// Indexed by gob types. tComplex will be added during type.init().
+var decIgnoreOpMap = map[typeId]decOp{
+ tBool: ignoreUint,
+ tInt: ignoreUint,
+ tUint: ignoreUint,
+ tFloat: ignoreUint,
+ tBytes: ignoreUint8Array,
+ tString: ignoreUint8Array,
+ tComplex: ignoreTwoUints,
+}
+
+// decOpFor returns the decoding op for the base type under rt and
+// the indirection count to reach it.
+func (dec *Decoder) decOpFor(wireId typeId, rt reflect.Type, name string, inProgress map[reflect.Type]*decOp) (*decOp, int) {
+ ut := userType(rt)
+ // If the type implements GobEncoder, we handle it without further processing.
+ if ut.isGobDecoder {
+ return dec.gobDecodeOpFor(ut)
+ }
+ // If this type is already in progress, it's a recursive type (e.g. map[string]*T).
+ // Return the pointer to the op we're already building.
+ if opPtr := inProgress[rt]; opPtr != nil {
+ return opPtr, ut.indir
+ }
+ typ := ut.base
+ indir := ut.indir
+ var op decOp
+ k := typ.Kind()
+ if int(k) < len(decOpTable) {
+ op = decOpTable[k]
+ }
+ if op == nil {
+ inProgress[rt] = &op
+ // Special cases
+ switch t := typ; t.Kind() {
+ case reflect.Array:
+ name = "element of " + name
+ elemId := dec.wireType[wireId].ArrayT.Elem
+ elemOp, elemIndir := dec.decOpFor(elemId, t.Elem(), name, inProgress)
+ ovfl := overflow(name)
+ op = func(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ state.dec.decodeArray(t, state, uintptr(p), *elemOp, t.Elem().Size(), t.Len(), i.indir, elemIndir, ovfl)
+ }
+
+ case reflect.Map:
+ name = "element of " + name
+ keyId := dec.wireType[wireId].MapT.Key
+ elemId := dec.wireType[wireId].MapT.Elem
+ keyOp, keyIndir := dec.decOpFor(keyId, t.Key(), name, inProgress)
+ elemOp, elemIndir := dec.decOpFor(elemId, t.Elem(), name, inProgress)
+ ovfl := overflow(name)
+ op = func(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ up := unsafe.Pointer(p)
+ state.dec.decodeMap(t, state, uintptr(up), *keyOp, *elemOp, i.indir, keyIndir, elemIndir, ovfl)
+ }
+
+ case reflect.Slice:
+ name = "element of " + name
+ if t.Elem().Kind() == reflect.Uint8 {
+ op = decUint8Slice
+ break
+ }
+ var elemId typeId
+ if tt, ok := builtinIdToType[wireId]; ok {
+ elemId = tt.(*sliceType).Elem
+ } else {
+ elemId = dec.wireType[wireId].SliceT.Elem
+ }
+ elemOp, elemIndir := dec.decOpFor(elemId, t.Elem(), name, inProgress)
+ ovfl := overflow(name)
+ op = func(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ state.dec.decodeSlice(t, state, uintptr(p), *elemOp, t.Elem().Size(), i.indir, elemIndir, ovfl)
+ }
+
+ case reflect.Struct:
+ // Generate a closure that calls out to the engine for the nested type.
+ enginePtr, err := dec.getDecEnginePtr(wireId, userType(typ))
+ if err != nil {
+ error_(err)
+ }
+ op = func(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ // indirect through enginePtr to delay evaluation for recursive structs.
+ dec.decodeStruct(*enginePtr, userType(typ), uintptr(p), i.indir)
+ }
+ case reflect.Interface:
+ op = func(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ state.dec.decodeInterface(t, state, uintptr(p), i.indir)
+ }
+ }
+ }
+ if op == nil {
+ errorf("decode can't handle type %s", rt)
+ }
+ return &op, indir
+}
+
+// decIgnoreOpFor returns the decoding op for a field that has no destination.
+func (dec *Decoder) decIgnoreOpFor(wireId typeId) decOp {
+ op, ok := decIgnoreOpMap[wireId]
+ if !ok {
+ if wireId == tInterface {
+ // Special case because it's a method: the ignored item might
+ // define types and we need to record their state in the decoder.
+ op = func(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ state.dec.ignoreInterface(state)
+ }
+ return op
+ }
+ // Special cases
+ wire := dec.wireType[wireId]
+ switch {
+ case wire == nil:
+ errorf("bad data: undefined type %s", wireId.string())
+ case wire.ArrayT != nil:
+ elemId := wire.ArrayT.Elem
+ elemOp := dec.decIgnoreOpFor(elemId)
+ op = func(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ state.dec.ignoreArray(state, elemOp, wire.ArrayT.Len)
+ }
+
+ case wire.MapT != nil:
+ keyId := dec.wireType[wireId].MapT.Key
+ elemId := dec.wireType[wireId].MapT.Elem
+ keyOp := dec.decIgnoreOpFor(keyId)
+ elemOp := dec.decIgnoreOpFor(elemId)
+ op = func(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ state.dec.ignoreMap(state, keyOp, elemOp)
+ }
+
+ case wire.SliceT != nil:
+ elemId := wire.SliceT.Elem
+ elemOp := dec.decIgnoreOpFor(elemId)
+ op = func(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ state.dec.ignoreSlice(state, elemOp)
+ }
+
+ case wire.StructT != nil:
+ // Generate a closure that calls out to the engine for the nested type.
+ enginePtr, err := dec.getIgnoreEnginePtr(wireId)
+ if err != nil {
+ error_(err)
+ }
+ op = func(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ // indirect through enginePtr to delay evaluation for recursive structs
+ state.dec.ignoreStruct(*enginePtr)
+ }
+
+ case wire.GobEncoderT != nil:
+ op = func(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ state.dec.ignoreGobDecoder(state)
+ }
+ }
+ }
+ if op == nil {
+ errorf("bad data: ignore can't handle type %s", wireId.string())
+ }
+ return op
+}
+
+// gobDecodeOpFor returns the op for a type that is known to implement
+// GobDecoder.
+func (dec *Decoder) gobDecodeOpFor(ut *userTypeInfo) (*decOp, int) {
+ rcvrType := ut.user
+ if ut.decIndir == -1 {
+ rcvrType = reflect.PtrTo(rcvrType)
+ } else if ut.decIndir > 0 {
+ for i := int8(0); i < ut.decIndir; i++ {
+ rcvrType = rcvrType.Elem()
+ }
+ }
+ var op decOp
+ op = func(i *decInstr, state *decoderState, p unsafe.Pointer) {
+ // Caller has gotten us to within one indirection of our value.
+ if i.indir > 0 {
+ if *(*unsafe.Pointer)(p) == nil {
+ *(*unsafe.Pointer)(p) = unsafe.New(ut.base)
+ }
+ }
+ // Now p is a pointer to the base type. Do we need to climb out to
+ // get to the receiver type?
+ var v reflect.Value
+ if ut.decIndir == -1 {
+ v = reflect.ValueOf(unsafe.Unreflect(rcvrType, unsafe.Pointer(&p)))
+ } else {
+ v = reflect.ValueOf(unsafe.Unreflect(rcvrType, p))
+ }
+ state.dec.decodeGobDecoder(state, v)
+ }
+ return &op, int(ut.indir)
+
+}
+
+// compatibleType asks: Are these two gob Types compatible?
+// Answers the question for basic types, arrays, maps and slices, plus
+// GobEncoder/Decoder pairs.
+// Structs are considered ok; fields will be checked later.
+func (dec *Decoder) compatibleType(fr reflect.Type, fw typeId, inProgress map[reflect.Type]typeId) bool {
+ if rhs, ok := inProgress[fr]; ok {
+ return rhs == fw
+ }
+ inProgress[fr] = fw
+ ut := userType(fr)
+ wire, ok := dec.wireType[fw]
+ // If fr is a GobDecoder, the wire type must be GobEncoder.
+ // And if fr is not a GobDecoder, the wire type must not be either.
+ if ut.isGobDecoder != (ok && wire.GobEncoderT != nil) { // the parentheses look odd but are correct.
+ return false
+ }
+ if ut.isGobDecoder { // This test trumps all others.
+ return true
+ }
+ switch t := ut.base; t.Kind() {
+ default:
+ // chan, etc: cannot handle.
+ return false
+ case reflect.Bool:
+ return fw == tBool
+ case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64:
+ return fw == tInt
+ case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr:
+ return fw == tUint
+ case reflect.Float32, reflect.Float64:
+ return fw == tFloat
+ case reflect.Complex64, reflect.Complex128:
+ return fw == tComplex
+ case reflect.String:
+ return fw == tString
+ case reflect.Interface:
+ return fw == tInterface
+ case reflect.Array:
+ if !ok || wire.ArrayT == nil {
+ return false
+ }
+ array := wire.ArrayT
+ return t.Len() == array.Len && dec.compatibleType(t.Elem(), array.Elem, inProgress)
+ case reflect.Map:
+ if !ok || wire.MapT == nil {
+ return false
+ }
+ MapType := wire.MapT
+ return dec.compatibleType(t.Key(), MapType.Key, inProgress) && dec.compatibleType(t.Elem(), MapType.Elem, inProgress)
+ case reflect.Slice:
+ // Is it an array of bytes?
+ if t.Elem().Kind() == reflect.Uint8 {
+ return fw == tBytes
+ }
+ // Extract and compare element types.
+ var sw *sliceType
+ if tt, ok := builtinIdToType[fw]; ok {
+ sw = tt.(*sliceType)
+ } else {
+ sw = dec.wireType[fw].SliceT
+ }
+ elem := userType(t.Elem()).base
+ return sw != nil && dec.compatibleType(elem, sw.Elem, inProgress)
+ case reflect.Struct:
+ return true
+ }
+ return true
+}
+
+// typeString returns a human-readable description of the type identified by remoteId.
+func (dec *Decoder) typeString(remoteId typeId) string {
+ if t := idToType[remoteId]; t != nil {
+ // globally known type.
+ return t.string()
+ }
+ return dec.wireType[remoteId].string()
+}
+
+// compileSingle compiles the decoder engine for a non-struct top-level value, including
+// GobDecoders.
+func (dec *Decoder) compileSingle(remoteId typeId, ut *userTypeInfo) (engine *decEngine, err error) {
+ rt := ut.user
+ engine = new(decEngine)
+ engine.instr = make([]decInstr, 1) // one item
+ name := rt.String() // best we can do
+ if !dec.compatibleType(rt, remoteId, make(map[reflect.Type]typeId)) {
+ return nil, errors.New("gob: wrong type received for local value " + name + ": " + dec.typeString(remoteId))
+ }
+ op, indir := dec.decOpFor(remoteId, rt, name, make(map[reflect.Type]*decOp))
+ ovfl := errors.New(`value for "` + name + `" out of range`)
+ engine.instr[singletonField] = decInstr{*op, singletonField, indir, 0, ovfl}
+ engine.numInstr = 1
+ return
+}
+
+// compileIgnoreSingle compiles the decoder engine for a non-struct top-level value that will be discarded.
+func (dec *Decoder) compileIgnoreSingle(remoteId typeId) (engine *decEngine, err error) {
+ engine = new(decEngine)
+ engine.instr = make([]decInstr, 1) // one item
+ op := dec.decIgnoreOpFor(remoteId)
+ ovfl := overflow(dec.typeString(remoteId))
+ engine.instr[0] = decInstr{op, 0, 0, 0, ovfl}
+ engine.numInstr = 1
+ return
+}
+
+// compileDec compiles the decoder engine for a value. If the value is not a struct,
+// it calls out to compileSingle.
+func (dec *Decoder) compileDec(remoteId typeId, ut *userTypeInfo) (engine *decEngine, err error) {
+ rt := ut.base
+ srt := rt
+ if srt.Kind() != reflect.Struct ||
+ ut.isGobDecoder {
+ return dec.compileSingle(remoteId, ut)
+ }
+ var wireStruct *structType
+ // Builtin types can come from global pool; the rest must be defined by the decoder.
+ // Also we know we're decoding a struct now, so the client must have sent one.
+ if t, ok := builtinIdToType[remoteId]; ok {
+ wireStruct, _ = t.(*structType)
+ } else {
+ wire := dec.wireType[remoteId]
+ if wire == nil {
+ error_(errBadType)
+ }
+ wireStruct = wire.StructT
+ }
+ if wireStruct == nil {
+ errorf("type mismatch in decoder: want struct type %s; got non-struct", rt)
+ }
+ engine = new(decEngine)
+ engine.instr = make([]decInstr, len(wireStruct.Field))
+ seen := make(map[reflect.Type]*decOp)
+ // Loop over the fields of the wire type.
+ for fieldnum := 0; fieldnum < len(wireStruct.Field); fieldnum++ {
+ wireField := wireStruct.Field[fieldnum]
+ if wireField.Name == "" {
+ errorf("empty name for remote field of type %s", wireStruct.Name)
+ }
+ ovfl := overflow(wireField.Name)
+ // Find the field of the local type with the same name.
+ localField, present := srt.FieldByName(wireField.Name)
+ // TODO(r): anonymous names
+ if !present || !isExported(wireField.Name) {
+ op := dec.decIgnoreOpFor(wireField.Id)
+ engine.instr[fieldnum] = decInstr{op, fieldnum, 0, 0, ovfl}
+ continue
+ }
+ if !dec.compatibleType(localField.Type, wireField.Id, make(map[reflect.Type]typeId)) {
+ errorf("wrong type (%s) for received field %s.%s", localField.Type, wireStruct.Name, wireField.Name)
+ }
+ op, indir := dec.decOpFor(wireField.Id, localField.Type, localField.Name, seen)
+ engine.instr[fieldnum] = decInstr{*op, fieldnum, indir, uintptr(localField.Offset), ovfl}
+ engine.numInstr++
+ }
+ return
+}
+
+// getDecEnginePtr returns the engine for the specified type.
+func (dec *Decoder) getDecEnginePtr(remoteId typeId, ut *userTypeInfo) (enginePtr **decEngine, err error) {
+ rt := ut.base
+ decoderMap, ok := dec.decoderCache[rt]
+ if !ok {
+ decoderMap = make(map[typeId]**decEngine)
+ dec.decoderCache[rt] = decoderMap
+ }
+ if enginePtr, ok = decoderMap[remoteId]; !ok {
+ // To handle recursive types, mark this engine as underway before compiling.
+ enginePtr = new(*decEngine)
+ decoderMap[remoteId] = enginePtr
+ *enginePtr, err = dec.compileDec(remoteId, ut)
+ if err != nil {
+ delete(decoderMap, remoteId)
+ }
+ }
+ return
+}
+
+// emptyStruct is the type we compile into when ignoring a struct value.
+type emptyStruct struct{}
+
+var emptyStructType = reflect.TypeOf(emptyStruct{})
+
+// getDecEnginePtr returns the engine for the specified type when the value is to be discarded.
+func (dec *Decoder) getIgnoreEnginePtr(wireId typeId) (enginePtr **decEngine, err error) {
+ var ok bool
+ if enginePtr, ok = dec.ignorerCache[wireId]; !ok {
+ // To handle recursive types, mark this engine as underway before compiling.
+ enginePtr = new(*decEngine)
+ dec.ignorerCache[wireId] = enginePtr
+ wire := dec.wireType[wireId]
+ if wire != nil && wire.StructT != nil {
+ *enginePtr, err = dec.compileDec(wireId, userType(emptyStructType))
+ } else {
+ *enginePtr, err = dec.compileIgnoreSingle(wireId)
+ }
+ if err != nil {
+ delete(dec.ignorerCache, wireId)
+ }
+ }
+ return
+}
+
+// decodeValue decodes the data stream representing a value and stores it in val.
+func (dec *Decoder) decodeValue(wireId typeId, val reflect.Value) {
+ defer catchError(&dec.err)
+ // If the value is nil, it means we should just ignore this item.
+ if !val.IsValid() {
+ dec.decodeIgnoredValue(wireId)
+ return
+ }
+ // Dereference down to the underlying type.
+ ut := userType(val.Type())
+ base := ut.base
+ var enginePtr **decEngine
+ enginePtr, dec.err = dec.getDecEnginePtr(wireId, ut)
+ if dec.err != nil {
+ return
+ }
+ engine := *enginePtr
+ if st := base; st.Kind() == reflect.Struct && !ut.isGobDecoder {
+ if engine.numInstr == 0 && st.NumField() > 0 && len(dec.wireType[wireId].StructT.Field) > 0 {
+ name := base.Name()
+ errorf("type mismatch: no fields matched compiling decoder for %s", name)
+ }
+ dec.decodeStruct(engine, ut, uintptr(unsafeAddr(val)), ut.indir)
+ } else {
+ dec.decodeSingle(engine, ut, uintptr(unsafeAddr(val)))
+ }
+}
+
+// decodeIgnoredValue decodes the data stream representing a value of the specified type and discards it.
+func (dec *Decoder) decodeIgnoredValue(wireId typeId) {
+ var enginePtr **decEngine
+ enginePtr, dec.err = dec.getIgnoreEnginePtr(wireId)
+ if dec.err != nil {
+ return
+ }
+ wire := dec.wireType[wireId]
+ if wire != nil && wire.StructT != nil {
+ dec.ignoreStruct(*enginePtr)
+ } else {
+ dec.ignoreSingle(*enginePtr)
+ }
+}
+
+func init() {
+ var iop, uop decOp
+ switch reflect.TypeOf(int(0)).Bits() {
+ case 32:
+ iop = decInt32
+ uop = decUint32
+ case 64:
+ iop = decInt64
+ uop = decUint64
+ default:
+ panic("gob: unknown size of int/uint")
+ }
+ decOpTable[reflect.Int] = iop
+ decOpTable[reflect.Uint] = uop
+
+ // Finally uintptr
+ switch reflect.TypeOf(uintptr(0)).Bits() {
+ case 32:
+ uop = decUint32
+ case 64:
+ uop = decUint64
+ default:
+ panic("gob: unknown size of uintptr")
+ }
+ decOpTable[reflect.Uintptr] = uop
+}
+
+// Gob assumes it can call UnsafeAddr on any Value
+// in order to get a pointer it can copy data from.
+// Values that have just been created and do not point
+// into existing structs or slices cannot be addressed,
+// so simulate it by returning a pointer to a copy.
+// Each call allocates once.
+func unsafeAddr(v reflect.Value) uintptr {
+ if v.CanAddr() {
+ return v.UnsafeAddr()
+ }
+ x := reflect.New(v.Type()).Elem()
+ x.Set(v)
+ return x.UnsafeAddr()
+}
+
+// Gob depends on being able to take the address
+// of zeroed Values it creates, so use this wrapper instead
+// of the standard reflect.Zero.
+// Each call allocates once.
+func allocValue(t reflect.Type) reflect.Value {
+ return reflect.New(t).Elem()
+}