Source file src/reflect/type.go

Documentation: reflect

  // 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 reflect implements run-time reflection, allowing a program to
  // manipulate objects with arbitrary types. The typical use is to take a value
  // with static type interface{} and extract its dynamic type information by
  // calling TypeOf, which returns a Type.
  //
  // A call to ValueOf returns a Value representing the run-time data.
  // Zero takes a Type and returns a Value representing a zero value
  // for that type.
  //
  // See "The Laws of Reflection" for an introduction to reflection in Go:
  // https://golang.org/doc/articles/laws_of_reflection.html
  package reflect
  
  import (
  	"runtime"
  	"strconv"
  	"sync"
  	"unicode"
  	"unicode/utf8"
  	"unsafe"
  )
  
  // Type is the representation of a Go type.
  //
  // Not all methods apply to all kinds of types. Restrictions,
  // if any, are noted in the documentation for each method.
  // Use the Kind method to find out the kind of type before
  // calling kind-specific methods. Calling a method
  // inappropriate to the kind of type causes a run-time panic.
  //
  // Type values are comparable, such as with the == operator,
  // so they can be used as map keys.
  // Two Type values are equal if they represent identical types.
  type Type interface {
  	// Methods applicable to all types.
  
  	// Align returns the alignment in bytes of a value of
  	// this type when allocated in memory.
  	Align() int
  
  	// FieldAlign returns the alignment in bytes of a value of
  	// this type when used as a field in a struct.
  	FieldAlign() int
  
  	// Method returns the i'th method in the type's method set.
  	// It panics if i is not in the range [0, NumMethod()).
  	//
  	// For a non-interface type T or *T, the returned Method's Type and Func
  	// fields describe a function whose first argument is the receiver.
  	//
  	// For an interface type, the returned Method's Type field gives the
  	// method signature, without a receiver, and the Func field is nil.
  	Method(int) Method
  
  	// MethodByName returns the method with that name in the type's
  	// method set and a boolean indicating if the method was found.
  	//
  	// For a non-interface type T or *T, the returned Method's Type and Func
  	// fields describe a function whose first argument is the receiver.
  	//
  	// For an interface type, the returned Method's Type field gives the
  	// method signature, without a receiver, and the Func field is nil.
  	MethodByName(string) (Method, bool)
  
  	// NumMethod returns the number of exported methods in the type's method set.
  	NumMethod() int
  
  	// Name returns the type's name within its package for a defined type.
  	// For other (non-defined) types it returns the empty string.
  	Name() string
  
  	// PkgPath returns a defined type's package path, that is, the import path
  	// that uniquely identifies the package, such as "encoding/base64".
  	// If the type was predeclared (string, error) or not defined (*T, struct{},
  	// []int, or A where A is an alias for a non-defined type), the package path
  	// will be the empty string.
  	PkgPath() string
  
  	// Size returns the number of bytes needed to store
  	// a value of the given type; it is analogous to unsafe.Sizeof.
  	Size() uintptr
  
  	// String returns a string representation of the type.
  	// The string representation may use shortened package names
  	// (e.g., base64 instead of "encoding/base64") and is not
  	// guaranteed to be unique among types. To test for type identity,
  	// compare the Types directly.
  	String() string
  
  	// Kind returns the specific kind of this type.
  	Kind() Kind
  
  	// Implements reports whether the type implements the interface type u.
  	Implements(u Type) bool
  
  	// AssignableTo reports whether a value of the type is assignable to type u.
  	AssignableTo(u Type) bool
  
  	// ConvertibleTo reports whether a value of the type is convertible to type u.
  	ConvertibleTo(u Type) bool
  
  	// Comparable reports whether values of this type are comparable.
  	Comparable() bool
  
  	// Methods applicable only to some types, depending on Kind.
  	// The methods allowed for each kind are:
  	//
  	//	Int*, Uint*, Float*, Complex*: Bits
  	//	Array: Elem, Len
  	//	Chan: ChanDir, Elem
  	//	Func: In, NumIn, Out, NumOut, IsVariadic.
  	//	Map: Key, Elem
  	//	Ptr: Elem
  	//	Slice: Elem
  	//	Struct: Field, FieldByIndex, FieldByName, FieldByNameFunc, NumField
  
  	// Bits returns the size of the type in bits.
  	// It panics if the type's Kind is not one of the
  	// sized or unsized Int, Uint, Float, or Complex kinds.
  	Bits() int
  
  	// ChanDir returns a channel type's direction.
  	// It panics if the type's Kind is not Chan.
  	ChanDir() ChanDir
  
  	// IsVariadic reports whether a function type's final input parameter
  	// is a "..." parameter. If so, t.In(t.NumIn() - 1) returns the parameter's
  	// implicit actual type []T.
  	//
  	// For concreteness, if t represents func(x int, y ... float64), then
  	//
  	//	t.NumIn() == 2
  	//	t.In(0) is the reflect.Type for "int"
  	//	t.In(1) is the reflect.Type for "[]float64"
  	//	t.IsVariadic() == true
  	//
  	// IsVariadic panics if the type's Kind is not Func.
  	IsVariadic() bool
  
  	// Elem returns a type's element type.
  	// It panics if the type's Kind is not Array, Chan, Map, Ptr, or Slice.
  	Elem() Type
  
  	// Field returns a struct type's i'th field.
  	// It panics if the type's Kind is not Struct.
  	// It panics if i is not in the range [0, NumField()).
  	Field(i int) StructField
  
  	// FieldByIndex returns the nested field corresponding
  	// to the index sequence. It is equivalent to calling Field
  	// successively for each index i.
  	// It panics if the type's Kind is not Struct.
  	FieldByIndex(index []int) StructField
  
  	// FieldByName returns the struct field with the given name
  	// and a boolean indicating if the field was found.
  	FieldByName(name string) (StructField, bool)
  
  	// FieldByNameFunc returns the struct field with a name
  	// that satisfies the match function and a boolean indicating if
  	// the field was found.
  	//
  	// FieldByNameFunc considers the fields in the struct itself
  	// and then the fields in any embedded structs, in breadth first order,
  	// stopping at the shallowest nesting depth containing one or more
  	// fields satisfying the match function. If multiple fields at that depth
  	// satisfy the match function, they cancel each other
  	// and FieldByNameFunc returns no match.
  	// This behavior mirrors Go's handling of name lookup in
  	// structs containing embedded fields.
  	FieldByNameFunc(match func(string) bool) (StructField, bool)
  
  	// In returns the type of a function type's i'th input parameter.
  	// It panics if the type's Kind is not Func.
  	// It panics if i is not in the range [0, NumIn()).
  	In(i int) Type
  
  	// Key returns a map type's key type.
  	// It panics if the type's Kind is not Map.
  	Key() Type
  
  	// Len returns an array type's length.
  	// It panics if the type's Kind is not Array.
  	Len() int
  
  	// NumField returns a struct type's field count.
  	// It panics if the type's Kind is not Struct.
  	NumField() int
  
  	// NumIn returns a function type's input parameter count.
  	// It panics if the type's Kind is not Func.
  	NumIn() int
  
  	// NumOut returns a function type's output parameter count.
  	// It panics if the type's Kind is not Func.
  	NumOut() int
  
  	// Out returns the type of a function type's i'th output parameter.
  	// It panics if the type's Kind is not Func.
  	// It panics if i is not in the range [0, NumOut()).
  	Out(i int) Type
  
  	common() *rtype
  	uncommon() *uncommonType
  }
  
  // BUG(rsc): FieldByName and related functions consider struct field names to be equal
  // if the names are equal, even if they are unexported names originating
  // in different packages. The practical effect of this is that the result of
  // t.FieldByName("x") is not well defined if the struct type t contains
  // multiple fields named x (embedded from different packages).
  // FieldByName may return one of the fields named x or may report that there are none.
  // See https://golang.org/issue/4876 for more details.
  
  /*
   * These data structures are known to the compiler (../../cmd/internal/gc/reflect.go).
   * A few are known to ../runtime/type.go to convey to debuggers.
   * They are also known to ../runtime/type.go.
   */
  
  // A Kind represents the specific kind of type that a Type represents.
  // The zero Kind is not a valid kind.
  type Kind uint
  
  const (
  	Invalid Kind = iota
  	Bool
  	Int
  	Int8
  	Int16
  	Int32
  	Int64
  	Uint
  	Uint8
  	Uint16
  	Uint32
  	Uint64
  	Uintptr
  	Float32
  	Float64
  	Complex64
  	Complex128
  	Array
  	Chan
  	Func
  	Interface
  	Map
  	Ptr
  	Slice
  	String
  	Struct
  	UnsafePointer
  )
  
  // tflag is used by an rtype to signal what extra type information is
  // available in the memory directly following the rtype value.
  //
  // tflag values must be kept in sync with copies in:
  //	cmd/compile/internal/gc/reflect.go
  //	cmd/link/internal/ld/decodesym.go
  //	runtime/type.go
  type tflag uint8
  
  const (
  	// tflagUncommon means that there is a pointer, *uncommonType,
  	// just beyond the outer type structure.
  	//
  	// For example, if t.Kind() == Struct and t.tflag&tflagUncommon != 0,
  	// then t has uncommonType data and it can be accessed as:
  	//
  	//	type tUncommon struct {
  	//		structType
  	//		u uncommonType
  	//	}
  	//	u := &(*tUncommon)(unsafe.Pointer(t)).u
  	tflagUncommon tflag = 1 << 0
  
  	// tflagExtraStar means the name in the str field has an
  	// extraneous '*' prefix. This is because for most types T in
  	// a program, the type *T also exists and reusing the str data
  	// saves binary size.
  	tflagExtraStar tflag = 1 << 1
  
  	// tflagNamed means the type has a name.
  	tflagNamed tflag = 1 << 2
  )
  
  // rtype is the common implementation of most values.
  // It is embedded in other struct types.
  //
  // rtype must be kept in sync with ../runtime/type.go:/^type._type.
  type rtype struct {
  	size       uintptr
  	ptrdata    uintptr  // number of bytes in the type that can contain pointers
  	hash       uint32   // hash of type; avoids computation in hash tables
  	tflag      tflag    // extra type information flags
  	align      uint8    // alignment of variable with this type
  	fieldAlign uint8    // alignment of struct field with this type
  	kind       uint8    // enumeration for C
  	alg        *typeAlg // algorithm table
  	gcdata     *byte    // garbage collection data
  	str        nameOff  // string form
  	ptrToThis  typeOff  // type for pointer to this type, may be zero
  }
  
  // a copy of runtime.typeAlg
  type typeAlg struct {
  	// function for hashing objects of this type
  	// (ptr to object, seed) -> hash
  	hash func(unsafe.Pointer, uintptr) uintptr
  	// function for comparing objects of this type
  	// (ptr to object A, ptr to object B) -> ==?
  	equal func(unsafe.Pointer, unsafe.Pointer) bool
  }
  
  // Method on non-interface type
  type method struct {
  	name nameOff // name of method
  	mtyp typeOff // method type (without receiver)
  	ifn  textOff // fn used in interface call (one-word receiver)
  	tfn  textOff // fn used for normal method call
  }
  
  // uncommonType is present only for defined types or types with methods
  // (if T is a defined type, the uncommonTypes for T and *T have methods).
  // Using a pointer to this struct reduces the overall size required
  // to describe a non-defined type with no methods.
  type uncommonType struct {
  	pkgPath nameOff // import path; empty for built-in types like int, string
  	mcount  uint16  // number of methods
  	xcount  uint16  // number of exported methods
  	moff    uint32  // offset from this uncommontype to [mcount]method
  	_       uint32  // unused
  }
  
  // ChanDir represents a channel type's direction.
  type ChanDir int
  
  const (
  	RecvDir ChanDir             = 1 << iota // <-chan
  	SendDir                                 // chan<-
  	BothDir = RecvDir | SendDir             // chan
  )
  
  // arrayType represents a fixed array type.
  type arrayType struct {
  	rtype
  	elem  *rtype // array element type
  	slice *rtype // slice type
  	len   uintptr
  }
  
  // chanType represents a channel type.
  type chanType struct {
  	rtype
  	elem *rtype  // channel element type
  	dir  uintptr // channel direction (ChanDir)
  }
  
  // funcType represents a function type.
  //
  // A *rtype for each in and out parameter is stored in an array that
  // directly follows the funcType (and possibly its uncommonType). So
  // a function type with one method, one input, and one output is:
  //
  //	struct {
  //		funcType
  //		uncommonType
  //		[2]*rtype    // [0] is in, [1] is out
  //	}
  type funcType struct {
  	rtype
  	inCount  uint16
  	outCount uint16 // top bit is set if last input parameter is ...
  }
  
  // imethod represents a method on an interface type
  type imethod struct {
  	name nameOff // name of method
  	typ  typeOff // .(*FuncType) underneath
  }
  
  // interfaceType represents an interface type.
  type interfaceType struct {
  	rtype
  	pkgPath name      // import path
  	methods []imethod // sorted by hash
  }
  
  // mapType represents a map type.
  type mapType struct {
  	rtype
  	key           *rtype // map key type
  	elem          *rtype // map element (value) type
  	bucket        *rtype // internal bucket structure
  	keysize       uint8  // size of key slot
  	indirectkey   uint8  // store ptr to key instead of key itself
  	valuesize     uint8  // size of value slot
  	indirectvalue uint8  // store ptr to value instead of value itself
  	bucketsize    uint16 // size of bucket
  	reflexivekey  bool   // true if k==k for all keys
  	needkeyupdate bool   // true if we need to update key on an overwrite
  }
  
  // ptrType represents a pointer type.
  type ptrType struct {
  	rtype
  	elem *rtype // pointer element (pointed at) type
  }
  
  // sliceType represents a slice type.
  type sliceType struct {
  	rtype
  	elem *rtype // slice element type
  }
  
  // Struct field
  type structField struct {
  	name        name    // name is always non-empty
  	typ         *rtype  // type of field
  	offsetEmbed uintptr // byte offset of field<<1 | isEmbedded
  }
  
  func (f *structField) offset() uintptr {
  	return f.offsetEmbed >> 1
  }
  
  func (f *structField) embedded() bool {
  	return f.offsetEmbed&1 != 0
  }
  
  // structType represents a struct type.
  type structType struct {
  	rtype
  	pkgPath name
  	fields  []structField // sorted by offset
  }
  
  // name is an encoded type name with optional extra data.
  //
  // The first byte is a bit field containing:
  //
  //	1<<0 the name is exported
  //	1<<1 tag data follows the name
  //	1<<2 pkgPath nameOff follows the name and tag
  //
  // The next two bytes are the data length:
  //
  //	 l := uint16(data[1])<<8 | uint16(data[2])
  //
  // Bytes [3:3+l] are the string data.
  //
  // If tag data follows then bytes 3+l and 3+l+1 are the tag length,
  // with the data following.
  //
  // If the import path follows, then 4 bytes at the end of
  // the data form a nameOff. The import path is only set for concrete
  // methods that are defined in a different package than their type.
  //
  // If a name starts with "*", then the exported bit represents
  // whether the pointed to type is exported.
  type name struct {
  	bytes *byte
  }
  
  func (n name) data(off int, whySafe string) *byte {
  	return (*byte)(add(unsafe.Pointer(n.bytes), uintptr(off), whySafe))
  }
  
  func (n name) isExported() bool {
  	return (*n.bytes)&(1<<0) != 0
  }
  
  func (n name) nameLen() int {
  	return int(uint16(*n.data(1, "name len field"))<<8 | uint16(*n.data(2, "name len field")))
  }
  
  func (n name) tagLen() int {
  	if *n.data(0, "name flag field")&(1<<1) == 0 {
  		return 0
  	}
  	off := 3 + n.nameLen()
  	return int(uint16(*n.data(off, "name taglen field"))<<8 | uint16(*n.data(off+1, "name taglen field")))
  }
  
  func (n name) name() (s string) {
  	if n.bytes == nil {
  		return
  	}
  	b := (*[4]byte)(unsafe.Pointer(n.bytes))
  
  	hdr := (*stringHeader)(unsafe.Pointer(&s))
  	hdr.Data = unsafe.Pointer(&b[3])
  	hdr.Len = int(b[1])<<8 | int(b[2])
  	return s
  }
  
  func (n name) tag() (s string) {
  	tl := n.tagLen()
  	if tl == 0 {
  		return ""
  	}
  	nl := n.nameLen()
  	hdr := (*stringHeader)(unsafe.Pointer(&s))
  	hdr.Data = unsafe.Pointer(n.data(3+nl+2, "non-empty string"))
  	hdr.Len = tl
  	return s
  }
  
  func (n name) pkgPath() string {
  	if n.bytes == nil || *n.data(0, "name flag field")&(1<<2) == 0 {
  		return ""
  	}
  	off := 3 + n.nameLen()
  	if tl := n.tagLen(); tl > 0 {
  		off += 2 + tl
  	}
  	var nameOff int32
  	// Note that this field may not be aligned in memory,
  	// so we cannot use a direct int32 assignment here.
  	copy((*[4]byte)(unsafe.Pointer(&nameOff))[:], (*[4]byte)(unsafe.Pointer(n.data(off, "name offset field")))[:])
  	pkgPathName := name{(*byte)(resolveTypeOff(unsafe.Pointer(n.bytes), nameOff))}
  	return pkgPathName.name()
  }
  
  // round n up to a multiple of a.  a must be a power of 2.
  func round(n, a uintptr) uintptr {
  	return (n + a - 1) &^ (a - 1)
  }
  
  func newName(n, tag string, exported bool) name {
  	if len(n) > 1<<16-1 {
  		panic("reflect.nameFrom: name too long: " + n)
  	}
  	if len(tag) > 1<<16-1 {
  		panic("reflect.nameFrom: tag too long: " + tag)
  	}
  
  	var bits byte
  	l := 1 + 2 + len(n)
  	if exported {
  		bits |= 1 << 0
  	}
  	if len(tag) > 0 {
  		l += 2 + len(tag)
  		bits |= 1 << 1
  	}
  
  	b := make([]byte, l)
  	b[0] = bits
  	b[1] = uint8(len(n) >> 8)
  	b[2] = uint8(len(n))
  	copy(b[3:], n)
  	if len(tag) > 0 {
  		tb := b[3+len(n):]
  		tb[0] = uint8(len(tag) >> 8)
  		tb[1] = uint8(len(tag))
  		copy(tb[2:], tag)
  	}
  
  	return name{bytes: &b[0]}
  }
  
  /*
   * The compiler knows the exact layout of all the data structures above.
   * The compiler does not know about the data structures and methods below.
   */
  
  // Method represents a single method.
  type Method struct {
  	// Name is the method name.
  	// PkgPath is the package path that qualifies a lower case (unexported)
  	// method name. It is empty for upper case (exported) method names.
  	// The combination of PkgPath and Name uniquely identifies a method
  	// in a method set.
  	// See https://golang.org/ref/spec#Uniqueness_of_identifiers
  	Name    string
  	PkgPath string
  
  	Type  Type  // method type
  	Func  Value // func with receiver as first argument
  	Index int   // index for Type.Method
  }
  
  const (
  	kindDirectIface = 1 << 5
  	kindGCProg      = 1 << 6 // Type.gc points to GC program
  	kindNoPointers  = 1 << 7
  	kindMask        = (1 << 5) - 1
  )
  
  func (k Kind) String() string {
  	if int(k) < len(kindNames) {
  		return kindNames[k]
  	}
  	return "kind" + strconv.Itoa(int(k))
  }
  
  var kindNames = []string{
  	Invalid:       "invalid",
  	Bool:          "bool",
  	Int:           "int",
  	Int8:          "int8",
  	Int16:         "int16",
  	Int32:         "int32",
  	Int64:         "int64",
  	Uint:          "uint",
  	Uint8:         "uint8",
  	Uint16:        "uint16",
  	Uint32:        "uint32",
  	Uint64:        "uint64",
  	Uintptr:       "uintptr",
  	Float32:       "float32",
  	Float64:       "float64",
  	Complex64:     "complex64",
  	Complex128:    "complex128",
  	Array:         "array",
  	Chan:          "chan",
  	Func:          "func",
  	Interface:     "interface",
  	Map:           "map",
  	Ptr:           "ptr",
  	Slice:         "slice",
  	String:        "string",
  	Struct:        "struct",
  	UnsafePointer: "unsafe.Pointer",
  }
  
  func (t *uncommonType) methods() []method {
  	if t.mcount == 0 {
  		return nil
  	}
  	return (*[1 << 16]method)(add(unsafe.Pointer(t), uintptr(t.moff), "t.mcount > 0"))[:t.mcount:t.mcount]
  }
  
  func (t *uncommonType) exportedMethods() []method {
  	if t.xcount == 0 {
  		return nil
  	}
  	return (*[1 << 16]method)(add(unsafe.Pointer(t), uintptr(t.moff), "t.xcount > 0"))[:t.xcount:t.xcount]
  }
  
  // resolveNameOff resolves a name offset from a base pointer.
  // The (*rtype).nameOff method is a convenience wrapper for this function.
  // Implemented in the runtime package.
  func resolveNameOff(ptrInModule unsafe.Pointer, off int32) unsafe.Pointer
  
  // resolveTypeOff resolves an *rtype offset from a base type.
  // The (*rtype).typeOff method is a convenience wrapper for this function.
  // Implemented in the runtime package.
  func resolveTypeOff(rtype unsafe.Pointer, off int32) unsafe.Pointer
  
  // resolveTextOff resolves an function pointer offset from a base type.
  // The (*rtype).textOff method is a convenience wrapper for this function.
  // Implemented in the runtime package.
  func resolveTextOff(rtype unsafe.Pointer, off int32) unsafe.Pointer
  
  // addReflectOff adds a pointer to the reflection lookup map in the runtime.
  // It returns a new ID that can be used as a typeOff or textOff, and will
  // be resolved correctly. Implemented in the runtime package.
  func addReflectOff(ptr unsafe.Pointer) int32
  
  // resolveReflectType adds a name to the reflection lookup map in the runtime.
  // It returns a new nameOff that can be used to refer to the pointer.
  func resolveReflectName(n name) nameOff {
  	return nameOff(addReflectOff(unsafe.Pointer(n.bytes)))
  }
  
  // resolveReflectType adds a *rtype to the reflection lookup map in the runtime.
  // It returns a new typeOff that can be used to refer to the pointer.
  func resolveReflectType(t *rtype) typeOff {
  	return typeOff(addReflectOff(unsafe.Pointer(t)))
  }
  
  // resolveReflectText adds a function pointer to the reflection lookup map in
  // the runtime. It returns a new textOff that can be used to refer to the
  // pointer.
  func resolveReflectText(ptr unsafe.Pointer) textOff {
  	return textOff(addReflectOff(ptr))
  }
  
  type nameOff int32 // offset to a name
  type typeOff int32 // offset to an *rtype
  type textOff int32 // offset from top of text section
  
  func (t *rtype) nameOff(off nameOff) name {
  	return name{(*byte)(resolveNameOff(unsafe.Pointer(t), int32(off)))}
  }
  
  func (t *rtype) typeOff(off typeOff) *rtype {
  	return (*rtype)(resolveTypeOff(unsafe.Pointer(t), int32(off)))
  }
  
  func (t *rtype) textOff(off textOff) unsafe.Pointer {
  	return resolveTextOff(unsafe.Pointer(t), int32(off))
  }
  
  func (t *rtype) uncommon() *uncommonType {
  	if t.tflag&tflagUncommon == 0 {
  		return nil
  	}
  	switch t.Kind() {
  	case Struct:
  		return &(*structTypeUncommon)(unsafe.Pointer(t)).u
  	case Ptr:
  		type u struct {
  			ptrType
  			u uncommonType
  		}
  		return &(*u)(unsafe.Pointer(t)).u
  	case Func:
  		type u struct {
  			funcType
  			u uncommonType
  		}
  		return &(*u)(unsafe.Pointer(t)).u
  	case Slice:
  		type u struct {
  			sliceType
  			u uncommonType
  		}
  		return &(*u)(unsafe.Pointer(t)).u
  	case Array:
  		type u struct {
  			arrayType
  			u uncommonType
  		}
  		return &(*u)(unsafe.Pointer(t)).u
  	case Chan:
  		type u struct {
  			chanType
  			u uncommonType
  		}
  		return &(*u)(unsafe.Pointer(t)).u
  	case Map:
  		type u struct {
  			mapType
  			u uncommonType
  		}
  		return &(*u)(unsafe.Pointer(t)).u
  	case Interface:
  		type u struct {
  			interfaceType
  			u uncommonType
  		}
  		return &(*u)(unsafe.Pointer(t)).u
  	default:
  		type u struct {
  			rtype
  			u uncommonType
  		}
  		return &(*u)(unsafe.Pointer(t)).u
  	}
  }
  
  func (t *rtype) String() string {
  	s := t.nameOff(t.str).name()
  	if t.tflag&tflagExtraStar != 0 {
  		return s[1:]
  	}
  	return s
  }
  
  func (t *rtype) Size() uintptr { return t.size }
  
  func (t *rtype) Bits() int {
  	if t == nil {
  		panic("reflect: Bits of nil Type")
  	}
  	k := t.Kind()
  	if k < Int || k > Complex128 {
  		panic("reflect: Bits of non-arithmetic Type " + t.String())
  	}
  	return int(t.size) * 8
  }
  
  func (t *rtype) Align() int { return int(t.align) }
  
  func (t *rtype) FieldAlign() int { return int(t.fieldAlign) }
  
  func (t *rtype) Kind() Kind { return Kind(t.kind & kindMask) }
  
  func (t *rtype) pointers() bool { return t.kind&kindNoPointers == 0 }
  
  func (t *rtype) common() *rtype { return t }
  
  func (t *rtype) exportedMethods() []method {
  	ut := t.uncommon()
  	if ut == nil {
  		return nil
  	}
  	return ut.exportedMethods()
  }
  
  func (t *rtype) NumMethod() int {
  	if t.Kind() == Interface {
  		tt := (*interfaceType)(unsafe.Pointer(t))
  		return tt.NumMethod()
  	}
  	return len(t.exportedMethods())
  }
  
  func (t *rtype) Method(i int) (m Method) {
  	if t.Kind() == Interface {
  		tt := (*interfaceType)(unsafe.Pointer(t))
  		return tt.Method(i)
  	}
  	methods := t.exportedMethods()
  	if i < 0 || i >= len(methods) {
  		panic("reflect: Method index out of range")
  	}
  	p := methods[i]
  	pname := t.nameOff(p.name)
  	m.Name = pname.name()
  	fl := flag(Func)
  	mtyp := t.typeOff(p.mtyp)
  	ft := (*funcType)(unsafe.Pointer(mtyp))
  	in := make([]Type, 0, 1+len(ft.in()))
  	in = append(in, t)
  	for _, arg := range ft.in() {
  		in = append(in, arg)
  	}
  	out := make([]Type, 0, len(ft.out()))
  	for _, ret := range ft.out() {
  		out = append(out, ret)
  	}
  	mt := FuncOf(in, out, ft.IsVariadic())
  	m.Type = mt
  	tfn := t.textOff(p.tfn)
  	fn := unsafe.Pointer(&tfn)
  	m.Func = Value{mt.(*rtype), fn, fl}
  
  	m.Index = i
  	return m
  }
  
  func (t *rtype) MethodByName(name string) (m Method, ok bool) {
  	if t.Kind() == Interface {
  		tt := (*interfaceType)(unsafe.Pointer(t))
  		return tt.MethodByName(name)
  	}
  	ut := t.uncommon()
  	if ut == nil {
  		return Method{}, false
  	}
  	// TODO(mdempsky): Binary search.
  	for i, p := range ut.exportedMethods() {
  		if t.nameOff(p.name).name() == name {
  			return t.Method(i), true
  		}
  	}
  	return Method{}, false
  }
  
  func (t *rtype) PkgPath() string {
  	if t.tflag&tflagNamed == 0 {
  		return ""
  	}
  	ut := t.uncommon()
  	if ut == nil {
  		return ""
  	}
  	return t.nameOff(ut.pkgPath).name()
  }
  
  func hasPrefix(s, prefix string) bool {
  	return len(s) >= len(prefix) && s[:len(prefix)] == prefix
  }
  
  func (t *rtype) Name() string {
  	if t.tflag&tflagNamed == 0 {
  		return ""
  	}
  	s := t.String()
  	i := len(s) - 1
  	for i >= 0 {
  		if s[i] == '.' {
  			break
  		}
  		i--
  	}
  	return s[i+1:]
  }
  
  func (t *rtype) ChanDir() ChanDir {
  	if t.Kind() != Chan {
  		panic("reflect: ChanDir of non-chan type")
  	}
  	tt := (*chanType)(unsafe.Pointer(t))
  	return ChanDir(tt.dir)
  }
  
  func (t *rtype) IsVariadic() bool {
  	if t.Kind() != Func {
  		panic("reflect: IsVariadic of non-func type")
  	}
  	tt := (*funcType)(unsafe.Pointer(t))
  	return tt.outCount&(1<<15) != 0
  }
  
  func (t *rtype) Elem() Type {
  	switch t.Kind() {
  	case Array:
  		tt := (*arrayType)(unsafe.Pointer(t))
  		return toType(tt.elem)
  	case Chan:
  		tt := (*chanType)(unsafe.Pointer(t))
  		return toType(tt.elem)
  	case Map:
  		tt := (*mapType)(unsafe.Pointer(t))
  		return toType(tt.elem)
  	case Ptr:
  		tt := (*ptrType)(unsafe.Pointer(t))
  		return toType(tt.elem)
  	case Slice:
  		tt := (*sliceType)(unsafe.Pointer(t))
  		return toType(tt.elem)
  	}
  	panic("reflect: Elem of invalid type")
  }
  
  func (t *rtype) Field(i int) StructField {
  	if t.Kind() != Struct {
  		panic("reflect: Field of non-struct type")
  	}
  	tt := (*structType)(unsafe.Pointer(t))
  	return tt.Field(i)
  }
  
  func (t *rtype) FieldByIndex(index []int) StructField {
  	if t.Kind() != Struct {
  		panic("reflect: FieldByIndex of non-struct type")
  	}
  	tt := (*structType)(unsafe.Pointer(t))
  	return tt.FieldByIndex(index)
  }
  
  func (t *rtype) FieldByName(name string) (StructField, bool) {
  	if t.Kind() != Struct {
  		panic("reflect: FieldByName of non-struct type")
  	}
  	tt := (*structType)(unsafe.Pointer(t))
  	return tt.FieldByName(name)
  }
  
  func (t *rtype) FieldByNameFunc(match func(string) bool) (StructField, bool) {
  	if t.Kind() != Struct {
  		panic("reflect: FieldByNameFunc of non-struct type")
  	}
  	tt := (*structType)(unsafe.Pointer(t))
  	return tt.FieldByNameFunc(match)
  }
  
  func (t *rtype) In(i int) Type {
  	if t.Kind() != Func {
  		panic("reflect: In of non-func type")
  	}
  	tt := (*funcType)(unsafe.Pointer(t))
  	return toType(tt.in()[i])
  }
  
  func (t *rtype) Key() Type {
  	if t.Kind() != Map {
  		panic("reflect: Key of non-map type")
  	}
  	tt := (*mapType)(unsafe.Pointer(t))
  	return toType(tt.key)
  }
  
  func (t *rtype) Len() int {
  	if t.Kind() != Array {
  		panic("reflect: Len of non-array type")
  	}
  	tt := (*arrayType)(unsafe.Pointer(t))
  	return int(tt.len)
  }
  
  func (t *rtype) NumField() int {
  	if t.Kind() != Struct {
  		panic("reflect: NumField of non-struct type")
  	}
  	tt := (*structType)(unsafe.Pointer(t))
  	return len(tt.fields)
  }
  
  func (t *rtype) NumIn() int {
  	if t.Kind() != Func {
  		panic("reflect: NumIn of non-func type")
  	}
  	tt := (*funcType)(unsafe.Pointer(t))
  	return int(tt.inCount)
  }
  
  func (t *rtype) NumOut() int {
  	if t.Kind() != Func {
  		panic("reflect: NumOut of non-func type")
  	}
  	tt := (*funcType)(unsafe.Pointer(t))
  	return len(tt.out())
  }
  
  func (t *rtype) Out(i int) Type {
  	if t.Kind() != Func {
  		panic("reflect: Out of non-func type")
  	}
  	tt := (*funcType)(unsafe.Pointer(t))
  	return toType(tt.out()[i])
  }
  
  func (t *funcType) in() []*rtype {
  	uadd := unsafe.Sizeof(*t)
  	if t.tflag&tflagUncommon != 0 {
  		uadd += unsafe.Sizeof(uncommonType{})
  	}
  	if t.inCount == 0 {
  		return nil
  	}
  	return (*[1 << 20]*rtype)(add(unsafe.Pointer(t), uadd, "t.inCount > 0"))[:t.inCount]
  }
  
  func (t *funcType) out() []*rtype {
  	uadd := unsafe.Sizeof(*t)
  	if t.tflag&tflagUncommon != 0 {
  		uadd += unsafe.Sizeof(uncommonType{})
  	}
  	outCount := t.outCount & (1<<15 - 1)
  	if outCount == 0 {
  		return nil
  	}
  	return (*[1 << 20]*rtype)(add(unsafe.Pointer(t), uadd, "outCount > 0"))[t.inCount : t.inCount+outCount]
  }
  
  // add returns p+x.
  //
  // The whySafe string is ignored, so that the function still inlines
  // as efficiently as p+x, but all call sites should use the string to
  // record why the addition is safe, which is to say why the addition
  // does not cause x to advance to the very end of p's allocation
  // and therefore point incorrectly at the next block in memory.
  func add(p unsafe.Pointer, x uintptr, whySafe string) unsafe.Pointer {
  	return unsafe.Pointer(uintptr(p) + x)
  }
  
  func (d ChanDir) String() string {
  	switch d {
  	case SendDir:
  		return "chan<-"
  	case RecvDir:
  		return "<-chan"
  	case BothDir:
  		return "chan"
  	}
  	return "ChanDir" + strconv.Itoa(int(d))
  }
  
  // Method returns the i'th method in the type's method set.
  func (t *interfaceType) Method(i int) (m Method) {
  	if i < 0 || i >= len(t.methods) {
  		return
  	}
  	p := &t.methods[i]
  	pname := t.nameOff(p.name)
  	m.Name = pname.name()
  	if !pname.isExported() {
  		m.PkgPath = pname.pkgPath()
  		if m.PkgPath == "" {
  			m.PkgPath = t.pkgPath.name()
  		}
  	}
  	m.Type = toType(t.typeOff(p.typ))
  	m.Index = i
  	return
  }
  
  // NumMethod returns the number of interface methods in the type's method set.
  func (t *interfaceType) NumMethod() int { return len(t.methods) }
  
  // MethodByName method with the given name in the type's method set.
  func (t *interfaceType) MethodByName(name string) (m Method, ok bool) {
  	if t == nil {
  		return
  	}
  	var p *imethod
  	for i := range t.methods {
  		p = &t.methods[i]
  		if t.nameOff(p.name).name() == name {
  			return t.Method(i), true
  		}
  	}
  	return
  }
  
  // A StructField describes a single field in a struct.
  type StructField struct {
  	// Name is the field name.
  	Name string
  	// PkgPath is the package path that qualifies a lower case (unexported)
  	// field name. It is empty for upper case (exported) field names.
  	// See https://golang.org/ref/spec#Uniqueness_of_identifiers
  	PkgPath string
  
  	Type      Type      // field type
  	Tag       StructTag // field tag string
  	Offset    uintptr   // offset within struct, in bytes
  	Index     []int     // index sequence for Type.FieldByIndex
  	Anonymous bool      // is an embedded field
  }
  
  // A StructTag is the tag string in a struct field.
  //
  // By convention, tag strings are a concatenation of
  // optionally space-separated key:"value" pairs.
  // Each key is a non-empty string consisting of non-control
  // characters other than space (U+0020 ' '), quote (U+0022 '"'),
  // and colon (U+003A ':').  Each value is quoted using U+0022 '"'
  // characters and Go string literal syntax.
  type StructTag string
  
  // Get returns the value associated with key in the tag string.
  // If there is no such key in the tag, Get returns the empty string.
  // If the tag does not have the conventional format, the value
  // returned by Get is unspecified. To determine whether a tag is
  // explicitly set to the empty string, use Lookup.
  func (tag StructTag) Get(key string) string {
  	v, _ := tag.Lookup(key)
  	return v
  }
  
  // Lookup returns the value associated with key in the tag string.
  // If the key is present in the tag the value (which may be empty)
  // is returned. Otherwise the returned value will be the empty string.
  // The ok return value reports whether the value was explicitly set in
  // the tag string. If the tag does not have the conventional format,
  // the value returned by Lookup is unspecified.
  func (tag StructTag) Lookup(key string) (value string, ok bool) {
  	// When modifying this code, also update the validateStructTag code
  	// in cmd/vet/structtag.go.
  
  	for tag != "" {
  		// Skip leading space.
  		i := 0
  		for i < len(tag) && tag[i] == ' ' {
  			i++
  		}
  		tag = tag[i:]
  		if tag == "" {
  			break
  		}
  
  		// Scan to colon. A space, a quote or a control character is a syntax error.
  		// Strictly speaking, control chars include the range [0x7f, 0x9f], not just
  		// [0x00, 0x1f], but in practice, we ignore the multi-byte control characters
  		// as it is simpler to inspect the tag's bytes than the tag's runes.
  		i = 0
  		for i < len(tag) && tag[i] > ' ' && tag[i] != ':' && tag[i] != '"' && tag[i] != 0x7f {
  			i++
  		}
  		if i == 0 || i+1 >= len(tag) || tag[i] != ':' || tag[i+1] != '"' {
  			break
  		}
  		name := string(tag[:i])
  		tag = tag[i+1:]
  
  		// Scan quoted string to find value.
  		i = 1
  		for i < len(tag) && tag[i] != '"' {
  			if tag[i] == '\\' {
  				i++
  			}
  			i++
  		}
  		if i >= len(tag) {
  			break
  		}
  		qvalue := string(tag[:i+1])
  		tag = tag[i+1:]
  
  		if key == name {
  			value, err := strconv.Unquote(qvalue)
  			if err != nil {
  				break
  			}
  			return value, true
  		}
  	}
  	return "", false
  }
  
  // Field returns the i'th struct field.
  func (t *structType) Field(i int) (f StructField) {
  	if i < 0 || i >= len(t.fields) {
  		panic("reflect: Field index out of bounds")
  	}
  	p := &t.fields[i]
  	f.Type = toType(p.typ)
  	f.Name = p.name.name()
  	f.Anonymous = p.embedded()
  	if !p.name.isExported() {
  		f.PkgPath = t.pkgPath.name()
  	}
  	if tag := p.name.tag(); tag != "" {
  		f.Tag = StructTag(tag)
  	}
  	f.Offset = p.offset()
  
  	// NOTE(rsc): This is the only allocation in the interface
  	// presented by a reflect.Type. It would be nice to avoid,
  	// at least in the common cases, but we need to make sure
  	// that misbehaving clients of reflect cannot affect other
  	// uses of reflect. One possibility is CL 5371098, but we
  	// postponed that ugliness until there is a demonstrated
  	// need for the performance. This is issue 2320.
  	f.Index = []int{i}
  	return
  }
  
  // TODO(gri): Should there be an error/bool indicator if the index
  //            is wrong for FieldByIndex?
  
  // FieldByIndex returns the nested field corresponding to index.
  func (t *structType) FieldByIndex(index []int) (f StructField) {
  	f.Type = toType(&t.rtype)
  	for i, x := range index {
  		if i > 0 {
  			ft := f.Type
  			if ft.Kind() == Ptr && ft.Elem().Kind() == Struct {
  				ft = ft.Elem()
  			}
  			f.Type = ft
  		}
  		f = f.Type.Field(x)
  	}
  	return
  }
  
  // A fieldScan represents an item on the fieldByNameFunc scan work list.
  type fieldScan struct {
  	typ   *structType
  	index []int
  }
  
  // FieldByNameFunc returns the struct field with a name that satisfies the
  // match function and a boolean to indicate if the field was found.
  func (t *structType) FieldByNameFunc(match func(string) bool) (result StructField, ok bool) {
  	// This uses the same condition that the Go language does: there must be a unique instance
  	// of the match at a given depth level. If there are multiple instances of a match at the
  	// same depth, they annihilate each other and inhibit any possible match at a lower level.
  	// The algorithm is breadth first search, one depth level at a time.
  
  	// The current and next slices are work queues:
  	// current lists the fields to visit on this depth level,
  	// and next lists the fields on the next lower level.
  	current := []fieldScan{}
  	next := []fieldScan{{typ: t}}
  
  	// nextCount records the number of times an embedded type has been
  	// encountered and considered for queueing in the 'next' slice.
  	// We only queue the first one, but we increment the count on each.
  	// If a struct type T can be reached more than once at a given depth level,
  	// then it annihilates itself and need not be considered at all when we
  	// process that next depth level.
  	var nextCount map[*structType]int
  
  	// visited records the structs that have been considered already.
  	// Embedded pointer fields can create cycles in the graph of
  	// reachable embedded types; visited avoids following those cycles.
  	// It also avoids duplicated effort: if we didn't find the field in an
  	// embedded type T at level 2, we won't find it in one at level 4 either.
  	visited := map[*structType]bool{}
  
  	for len(next) > 0 {
  		current, next = next, current[:0]
  		count := nextCount
  		nextCount = nil
  
  		// Process all the fields at this depth, now listed in 'current'.
  		// The loop queues embedded fields found in 'next', for processing during the next
  		// iteration. The multiplicity of the 'current' field counts is recorded
  		// in 'count'; the multiplicity of the 'next' field counts is recorded in 'nextCount'.
  		for _, scan := range current {
  			t := scan.typ
  			if visited[t] {
  				// We've looked through this type before, at a higher level.
  				// That higher level would shadow the lower level we're now at,
  				// so this one can't be useful to us. Ignore it.
  				continue
  			}
  			visited[t] = true
  			for i := range t.fields {
  				f := &t.fields[i]
  				// Find name and (for embedded field) type for field f.
  				fname := f.name.name()
  				var ntyp *rtype
  				if f.embedded() {
  					// Embedded field of type T or *T.
  					ntyp = f.typ
  					if ntyp.Kind() == Ptr {
  						ntyp = ntyp.Elem().common()
  					}
  				}
  
  				// Does it match?
  				if match(fname) {
  					// Potential match
  					if count[t] > 1 || ok {
  						// Name appeared multiple times at this level: annihilate.
  						return StructField{}, false
  					}
  					result = t.Field(i)
  					result.Index = nil
  					result.Index = append(result.Index, scan.index...)
  					result.Index = append(result.Index, i)
  					ok = true
  					continue
  				}
  
  				// Queue embedded struct fields for processing with next level,
  				// but only if we haven't seen a match yet at this level and only
  				// if the embedded types haven't already been queued.
  				if ok || ntyp == nil || ntyp.Kind() != Struct {
  					continue
  				}
  				styp := (*structType)(unsafe.Pointer(ntyp))
  				if nextCount[styp] > 0 {
  					nextCount[styp] = 2 // exact multiple doesn't matter
  					continue
  				}
  				if nextCount == nil {
  					nextCount = map[*structType]int{}
  				}
  				nextCount[styp] = 1
  				if count[t] > 1 {
  					nextCount[styp] = 2 // exact multiple doesn't matter
  				}
  				var index []int
  				index = append(index, scan.index...)
  				index = append(index, i)
  				next = append(next, fieldScan{styp, index})
  			}
  		}
  		if ok {
  			break
  		}
  	}
  	return
  }
  
  // FieldByName returns the struct field with the given name
  // and a boolean to indicate if the field was found.
  func (t *structType) FieldByName(name string) (f StructField, present bool) {
  	// Quick check for top-level name, or struct without embedded fields.
  	hasEmbeds := false
  	if name != "" {
  		for i := range t.fields {
  			tf := &t.fields[i]
  			if tf.name.name() == name {
  				return t.Field(i), true
  			}
  			if tf.embedded() {
  				hasEmbeds = true
  			}
  		}
  	}
  	if !hasEmbeds {
  		return
  	}
  	return t.FieldByNameFunc(func(s string) bool { return s == name })
  }
  
  // TypeOf returns the reflection Type that represents the dynamic type of i.
  // If i is a nil interface value, TypeOf returns nil.
  func TypeOf(i interface{}) Type {
  	eface := *(*emptyInterface)(unsafe.Pointer(&i))
  	return toType(eface.typ)
  }
  
  // ptrMap is the cache for PtrTo.
  var ptrMap sync.Map // map[*rtype]*ptrType
  
  // PtrTo returns the pointer type with element t.
  // For example, if t represents type Foo, PtrTo(t) represents *Foo.
  func PtrTo(t Type) Type {
  	return t.(*rtype).ptrTo()
  }
  
  func (t *rtype) ptrTo() *rtype {
  	if t.ptrToThis != 0 {
  		return t.typeOff(t.ptrToThis)
  	}
  
  	// Check the cache.
  	if pi, ok := ptrMap.Load(t); ok {
  		return &pi.(*ptrType).rtype
  	}
  
  	// Look in known types.
  	s := "*" + t.String()
  	for _, tt := range typesByString(s) {
  		p := (*ptrType)(unsafe.Pointer(tt))
  		if p.elem != t {
  			continue
  		}
  		pi, _ := ptrMap.LoadOrStore(t, p)
  		return &pi.(*ptrType).rtype
  	}
  
  	// Create a new ptrType starting with the description
  	// of an *unsafe.Pointer.
  	var iptr interface{} = (*unsafe.Pointer)(nil)
  	prototype := *(**ptrType)(unsafe.Pointer(&iptr))
  	pp := *prototype
  
  	pp.str = resolveReflectName(newName(s, "", false))
  	pp.ptrToThis = 0
  
  	// For the type structures linked into the binary, the
  	// compiler provides a good hash of the string.
  	// Create a good hash for the new string by using
  	// the FNV-1 hash's mixing function to combine the
  	// old hash and the new "*".
  	pp.hash = fnv1(t.hash, '*')
  
  	pp.elem = t
  
  	pi, _ := ptrMap.LoadOrStore(t, &pp)
  	return &pi.(*ptrType).rtype
  }
  
  // fnv1 incorporates the list of bytes into the hash x using the FNV-1 hash function.
  func fnv1(x uint32, list ...byte) uint32 {
  	for _, b := range list {
  		x = x*16777619 ^ uint32(b)
  	}
  	return x
  }
  
  func (t *rtype) Implements(u Type) bool {
  	if u == nil {
  		panic("reflect: nil type passed to Type.Implements")
  	}
  	if u.Kind() != Interface {
  		panic("reflect: non-interface type passed to Type.Implements")
  	}
  	return implements(u.(*rtype), t)
  }
  
  func (t *rtype) AssignableTo(u Type) bool {
  	if u == nil {
  		panic("reflect: nil type passed to Type.AssignableTo")
  	}
  	uu := u.(*rtype)
  	return directlyAssignable(uu, t) || implements(uu, t)
  }
  
  func (t *rtype) ConvertibleTo(u Type) bool {
  	if u == nil {
  		panic("reflect: nil type passed to Type.ConvertibleTo")
  	}
  	uu := u.(*rtype)
  	return convertOp(uu, t) != nil
  }
  
  func (t *rtype) Comparable() bool {
  	return t.alg != nil && t.alg.equal != nil
  }
  
  // implements reports whether the type V implements the interface type T.
  func implements(T, V *rtype) bool {
  	if T.Kind() != Interface {
  		return false
  	}
  	t := (*interfaceType)(unsafe.Pointer(T))
  	if len(t.methods) == 0 {
  		return true
  	}
  
  	// The same algorithm applies in both cases, but the
  	// method tables for an interface type and a concrete type
  	// are different, so the code is duplicated.
  	// In both cases the algorithm is a linear scan over the two
  	// lists - T's methods and V's methods - simultaneously.
  	// Since method tables are stored in a unique sorted order
  	// (alphabetical, with no duplicate method names), the scan
  	// through V's methods must hit a match for each of T's
  	// methods along the way, or else V does not implement T.
  	// This lets us run the scan in overall linear time instead of
  	// the quadratic time  a naive search would require.
  	// See also ../runtime/iface.go.
  	if V.Kind() == Interface {
  		v := (*interfaceType)(unsafe.Pointer(V))
  		i := 0
  		for j := 0; j < len(v.methods); j++ {
  			tm := &t.methods[i]
  			tmName := t.nameOff(tm.name)
  			vm := &v.methods[j]
  			vmName := V.nameOff(vm.name)
  			if vmName.name() == tmName.name() && V.typeOff(vm.typ) == t.typeOff(tm.typ) {
  				if !tmName.isExported() {
  					tmPkgPath := tmName.pkgPath()
  					if tmPkgPath == "" {
  						tmPkgPath = t.pkgPath.name()
  					}
  					vmPkgPath := vmName.pkgPath()
  					if vmPkgPath == "" {
  						vmPkgPath = v.pkgPath.name()
  					}
  					if tmPkgPath != vmPkgPath {
  						continue
  					}
  				}
  				if i++; i >= len(t.methods) {
  					return true
  				}
  			}
  		}
  		return false
  	}
  
  	v := V.uncommon()
  	if v == nil {
  		return false
  	}
  	i := 0
  	vmethods := v.methods()
  	for j := 0; j < int(v.mcount); j++ {
  		tm := &t.methods[i]
  		tmName := t.nameOff(tm.name)
  		vm := vmethods[j]
  		vmName := V.nameOff(vm.name)
  		if vmName.name() == tmName.name() && V.typeOff(vm.mtyp) == t.typeOff(tm.typ) {
  			if !tmName.isExported() {
  				tmPkgPath := tmName.pkgPath()
  				if tmPkgPath == "" {
  					tmPkgPath = t.pkgPath.name()
  				}
  				vmPkgPath := vmName.pkgPath()
  				if vmPkgPath == "" {
  					vmPkgPath = V.nameOff(v.pkgPath).name()
  				}
  				if tmPkgPath != vmPkgPath {
  					continue
  				}
  			}
  			if i++; i >= len(t.methods) {
  				return true
  			}
  		}
  	}
  	return false
  }
  
  // directlyAssignable reports whether a value x of type V can be directly
  // assigned (using memmove) to a value of type T.
  // https://golang.org/doc/go_spec.html#Assignability
  // Ignoring the interface rules (implemented elsewhere)
  // and the ideal constant rules (no ideal constants at run time).
  func directlyAssignable(T, V *rtype) bool {
  	// x's type V is identical to T?
  	if T == V {
  		return true
  	}
  
  	// Otherwise at least one of T and V must not be defined
  	// and they must have the same kind.
  	if T.Name() != "" && V.Name() != "" || T.Kind() != V.Kind() {
  		return false
  	}
  
  	// x's type T and V must  have identical underlying types.
  	return haveIdenticalUnderlyingType(T, V, true)
  }
  
  func haveIdenticalType(T, V Type, cmpTags bool) bool {
  	if cmpTags {
  		return T == V
  	}
  
  	if T.Name() != V.Name() || T.Kind() != V.Kind() {
  		return false
  	}
  
  	return haveIdenticalUnderlyingType(T.common(), V.common(), false)
  }
  
  func haveIdenticalUnderlyingType(T, V *rtype, cmpTags bool) bool {
  	if T == V {
  		return true
  	}
  
  	kind := T.Kind()
  	if kind != V.Kind() {
  		return false
  	}
  
  	// Non-composite types of equal kind have same underlying type
  	// (the predefined instance of the type).
  	if Bool <= kind && kind <= Complex128 || kind == String || kind == UnsafePointer {
  		return true
  	}
  
  	// Composite types.
  	switch kind {
  	case Array:
  		return T.Len() == V.Len() && haveIdenticalType(T.Elem(), V.Elem(), cmpTags)
  
  	case Chan:
  		// Special case:
  		// x is a bidirectional channel value, T is a channel type,
  		// and x's type V and T have identical element types.
  		if V.ChanDir() == BothDir && haveIdenticalType(T.Elem(), V.Elem(), cmpTags) {
  			return true
  		}
  
  		// Otherwise continue test for identical underlying type.
  		return V.ChanDir() == T.ChanDir() && haveIdenticalType(T.Elem(), V.Elem(), cmpTags)
  
  	case Func:
  		t := (*funcType)(unsafe.Pointer(T))
  		v := (*funcType)(unsafe.Pointer(V))
  		if t.outCount != v.outCount || t.inCount != v.inCount {
  			return false
  		}
  		for i := 0; i < t.NumIn(); i++ {
  			if !haveIdenticalType(t.In(i), v.In(i), cmpTags) {
  				return false
  			}
  		}
  		for i := 0; i < t.NumOut(); i++ {
  			if !haveIdenticalType(t.Out(i), v.Out(i), cmpTags) {
  				return false
  			}
  		}
  		return true
  
  	case Interface:
  		t := (*interfaceType)(unsafe.Pointer(T))
  		v := (*interfaceType)(unsafe.Pointer(V))
  		if len(t.methods) == 0 && len(v.methods) == 0 {
  			return true
  		}
  		// Might have the same methods but still
  		// need a run time conversion.
  		return false
  
  	case Map:
  		return haveIdenticalType(T.Key(), V.Key(), cmpTags) && haveIdenticalType(T.Elem(), V.Elem(), cmpTags)
  
  	case Ptr, Slice:
  		return haveIdenticalType(T.Elem(), V.Elem(), cmpTags)
  
  	case Struct:
  		t := (*structType)(unsafe.Pointer(T))
  		v := (*structType)(unsafe.Pointer(V))
  		if len(t.fields) != len(v.fields) {
  			return false
  		}
  		if t.pkgPath.name() != v.pkgPath.name() {
  			return false
  		}
  		for i := range t.fields {
  			tf := &t.fields[i]
  			vf := &v.fields[i]
  			if tf.name.name() != vf.name.name() {
  				return false
  			}
  			if !haveIdenticalType(tf.typ, vf.typ, cmpTags) {
  				return false
  			}
  			if cmpTags && tf.name.tag() != vf.name.tag() {
  				return false
  			}
  			if tf.offsetEmbed != vf.offsetEmbed {
  				return false
  			}
  		}
  		return true
  	}
  
  	return false
  }
  
  // typelinks is implemented in package runtime.
  // It returns a slice of the sections in each module,
  // and a slice of *rtype offsets in each module.
  //
  // The types in each module are sorted by string. That is, the first
  // two linked types of the first module are:
  //
  //	d0 := sections[0]
  //	t1 := (*rtype)(add(d0, offset[0][0]))
  //	t2 := (*rtype)(add(d0, offset[0][1]))
  //
  // and
  //
  //	t1.String() < t2.String()
  //
  // Note that strings are not unique identifiers for types:
  // there can be more than one with a given string.
  // Only types we might want to look up are included:
  // pointers, channels, maps, slices, and arrays.
  func typelinks() (sections []unsafe.Pointer, offset [][]int32)
  
  func rtypeOff(section unsafe.Pointer, off int32) *rtype {
  	return (*rtype)(add(section, uintptr(off), "sizeof(rtype) > 0"))
  }
  
  // typesByString returns the subslice of typelinks() whose elements have
  // the given string representation.
  // It may be empty (no known types with that string) or may have
  // multiple elements (multiple types with that string).
  func typesByString(s string) []*rtype {
  	sections, offset := typelinks()
  	var ret []*rtype
  
  	for offsI, offs := range offset {
  		section := sections[offsI]
  
  		// We are looking for the first index i where the string becomes >= s.
  		// This is a copy of sort.Search, with f(h) replaced by (*typ[h].String() >= s).
  		i, j := 0, len(offs)
  		for i < j {
  			h := i + (j-i)/2 // avoid overflow when computing h
  			// i ≤ h < j
  			if !(rtypeOff(section, offs[h]).String() >= s) {
  				i = h + 1 // preserves f(i-1) == false
  			} else {
  				j = h // preserves f(j) == true
  			}
  		}
  		// i == j, f(i-1) == false, and f(j) (= f(i)) == true  =>  answer is i.
  
  		// Having found the first, linear scan forward to find the last.
  		// We could do a second binary search, but the caller is going
  		// to do a linear scan anyway.
  		for j := i; j < len(offs); j++ {
  			typ := rtypeOff(section, offs[j])
  			if typ.String() != s {
  				break
  			}
  			ret = append(ret, typ)
  		}
  	}
  	return ret
  }
  
  // The lookupCache caches ArrayOf, ChanOf, MapOf and SliceOf lookups.
  var lookupCache sync.Map // map[cacheKey]*rtype
  
  // A cacheKey is the key for use in the lookupCache.
  // Four values describe any of the types we are looking for:
  // type kind, one or two subtypes, and an extra integer.
  type cacheKey struct {
  	kind  Kind
  	t1    *rtype
  	t2    *rtype
  	extra uintptr
  }
  
  // The funcLookupCache caches FuncOf lookups.
  // FuncOf does not share the common lookupCache since cacheKey is not
  // sufficient to represent functions unambiguously.
  var funcLookupCache struct {
  	sync.Mutex // Guards stores (but not loads) on m.
  
  	// m is a map[uint32][]*rtype keyed by the hash calculated in FuncOf.
  	// Elements of m are append-only and thus safe for concurrent reading.
  	m sync.Map
  }
  
  // ChanOf returns the channel type with the given direction and element type.
  // For example, if t represents int, ChanOf(RecvDir, t) represents <-chan int.
  //
  // The gc runtime imposes a limit of 64 kB on channel element types.
  // If t's size is equal to or exceeds this limit, ChanOf panics.
  func ChanOf(dir ChanDir, t Type) Type {
  	typ := t.(*rtype)
  
  	// Look in cache.
  	ckey := cacheKey{Chan, typ, nil, uintptr(dir)}
  	if ch, ok := lookupCache.Load(ckey); ok {
  		return ch.(*rtype)
  	}
  
  	// This restriction is imposed by the gc compiler and the runtime.
  	if typ.size >= 1<<16 {
  		panic("reflect.ChanOf: element size too large")
  	}
  
  	// Look in known types.
  	// TODO: Precedence when constructing string.
  	var s string
  	switch dir {
  	default:
  		panic("reflect.ChanOf: invalid dir")
  	case SendDir:
  		s = "chan<- " + typ.String()
  	case RecvDir:
  		s = "<-chan " + typ.String()
  	case BothDir:
  		s = "chan " + typ.String()
  	}
  	for _, tt := range typesByString(s) {
  		ch := (*chanType)(unsafe.Pointer(tt))
  		if ch.elem == typ && ch.dir == uintptr(dir) {
  			ti, _ := lookupCache.LoadOrStore(ckey, tt)
  			return ti.(Type)
  		}
  	}
  
  	// Make a channel type.
  	var ichan interface{} = (chan unsafe.Pointer)(nil)
  	prototype := *(**chanType)(unsafe.Pointer(&ichan))
  	ch := *prototype
  	ch.tflag = 0
  	ch.dir = uintptr(dir)
  	ch.str = resolveReflectName(newName(s, "", false))
  	ch.hash = fnv1(typ.hash, 'c', byte(dir))
  	ch.elem = typ
  
  	ti, _ := lookupCache.LoadOrStore(ckey, &ch.rtype)
  	return ti.(Type)
  }
  
  func ismapkey(*rtype) bool // implemented in runtime
  
  // MapOf returns the map type with the given key and element types.
  // For example, if k represents int and e represents string,
  // MapOf(k, e) represents map[int]string.
  //
  // If the key type is not a valid map key type (that is, if it does
  // not implement Go's == operator), MapOf panics.
  func MapOf(key, elem Type) Type {
  	ktyp := key.(*rtype)
  	etyp := elem.(*rtype)
  
  	if !ismapkey(ktyp) {
  		panic("reflect.MapOf: invalid key type " + ktyp.String())
  	}
  
  	// Look in cache.
  	ckey := cacheKey{Map, ktyp, etyp, 0}
  	if mt, ok := lookupCache.Load(ckey); ok {
  		return mt.(Type)
  	}
  
  	// Look in known types.
  	s := "map[" + ktyp.String() + "]" + etyp.String()
  	for _, tt := range typesByString(s) {
  		mt := (*mapType)(unsafe.Pointer(tt))
  		if mt.key == ktyp && mt.elem == etyp {
  			ti, _ := lookupCache.LoadOrStore(ckey, tt)
  			return ti.(Type)
  		}
  	}
  
  	// Make a map type.
  	var imap interface{} = (map[unsafe.Pointer]unsafe.Pointer)(nil)
  	mt := **(**mapType)(unsafe.Pointer(&imap))
  	mt.str = resolveReflectName(newName(s, "", false))
  	mt.tflag = 0
  	mt.hash = fnv1(etyp.hash, 'm', byte(ktyp.hash>>24), byte(ktyp.hash>>16), byte(ktyp.hash>>8), byte(ktyp.hash))
  	mt.key = ktyp
  	mt.elem = etyp
  	mt.bucket = bucketOf(ktyp, etyp)
  	if ktyp.size > maxKeySize {
  		mt.keysize = uint8(ptrSize)
  		mt.indirectkey = 1
  	} else {
  		mt.keysize = uint8(ktyp.size)
  		mt.indirectkey = 0
  	}
  	if etyp.size > maxValSize {
  		mt.valuesize = uint8(ptrSize)
  		mt.indirectvalue = 1
  	} else {
  		mt.valuesize = uint8(etyp.size)
  		mt.indirectvalue = 0
  	}
  	mt.bucketsize = uint16(mt.bucket.size)
  	mt.reflexivekey = isReflexive(ktyp)
  	mt.needkeyupdate = needKeyUpdate(ktyp)
  	mt.ptrToThis = 0
  
  	ti, _ := lookupCache.LoadOrStore(ckey, &mt.rtype)
  	return ti.(Type)
  }
  
  type funcTypeFixed4 struct {
  	funcType
  	args [4]*rtype
  }
  type funcTypeFixed8 struct {
  	funcType
  	args [8]*rtype
  }
  type funcTypeFixed16 struct {
  	funcType
  	args [16]*rtype
  }
  type funcTypeFixed32 struct {
  	funcType
  	args [32]*rtype
  }
  type funcTypeFixed64 struct {
  	funcType
  	args [64]*rtype
  }
  type funcTypeFixed128 struct {
  	funcType
  	args [128]*rtype
  }
  
  // FuncOf returns the function type with the given argument and result types.
  // For example if k represents int and e represents string,
  // FuncOf([]Type{k}, []Type{e}, false) represents func(int) string.
  //
  // The variadic argument controls whether the function is variadic. FuncOf
  // panics if the in[len(in)-1] does not represent a slice and variadic is
  // true.
  func FuncOf(in, out []Type, variadic bool) Type {
  	if variadic && (len(in) == 0 || in[len(in)-1].Kind() != Slice) {
  		panic("reflect.FuncOf: last arg of variadic func must be slice")
  	}
  
  	// Make a func type.
  	var ifunc interface{} = (func())(nil)
  	prototype := *(**funcType)(unsafe.Pointer(&ifunc))
  	n := len(in) + len(out)
  
  	var ft *funcType
  	var args []*rtype
  	switch {
  	case n <= 4:
  		fixed := new(funcTypeFixed4)
  		args = fixed.args[:0:len(fixed.args)]
  		ft = &fixed.funcType
  	case n <= 8:
  		fixed := new(funcTypeFixed8)
  		args = fixed.args[:0:len(fixed.args)]
  		ft = &fixed.funcType
  	case n <= 16:
  		fixed := new(funcTypeFixed16)
  		args = fixed.args[:0:len(fixed.args)]
  		ft = &fixed.funcType
  	case n <= 32:
  		fixed := new(funcTypeFixed32)
  		args = fixed.args[:0:len(fixed.args)]
  		ft = &fixed.funcType
  	case n <= 64:
  		fixed := new(funcTypeFixed64)
  		args = fixed.args[:0:len(fixed.args)]
  		ft = &fixed.funcType
  	case n <= 128:
  		fixed := new(funcTypeFixed128)
  		args = fixed.args[:0:len(fixed.args)]
  		ft = &fixed.funcType
  	default:
  		panic("reflect.FuncOf: too many arguments")
  	}
  	*ft = *prototype
  
  	// Build a hash and minimally populate ft.
  	var hash uint32
  	for _, in := range in {
  		t := in.(*rtype)
  		args = append(args, t)
  		hash = fnv1(hash, byte(t.hash>>24), byte(t.hash>>16), byte(t.hash>>8), byte(t.hash))
  	}
  	if variadic {
  		hash = fnv1(hash, 'v')
  	}
  	hash = fnv1(hash, '.')
  	for _, out := range out {
  		t := out.(*rtype)
  		args = append(args, t)
  		hash = fnv1(hash, byte(t.hash>>24), byte(t.hash>>16), byte(t.hash>>8), byte(t.hash))
  	}
  	if len(args) > 50 {
  		panic("reflect.FuncOf does not support more than 50 arguments")
  	}
  	ft.tflag = 0
  	ft.hash = hash
  	ft.inCount = uint16(len(in))
  	ft.outCount = uint16(len(out))
  	if variadic {
  		ft.outCount |= 1 << 15
  	}
  
  	// Look in cache.
  	if ts, ok := funcLookupCache.m.Load(hash); ok {
  		for _, t := range ts.([]*rtype) {
  			if haveIdenticalUnderlyingType(&ft.rtype, t, true) {
  				return t
  			}
  		}
  	}
  
  	// Not in cache, lock and retry.
  	funcLookupCache.Lock()
  	defer funcLookupCache.Unlock()
  	if ts, ok := funcLookupCache.m.Load(hash); ok {
  		for _, t := range ts.([]*rtype) {
  			if haveIdenticalUnderlyingType(&ft.rtype, t, true) {
  				return t
  			}
  		}
  	}
  
  	addToCache := func(tt *rtype) Type {
  		var rts []*rtype
  		if rti, ok := funcLookupCache.m.Load(hash); ok {
  			rts = rti.([]*rtype)
  		}
  		funcLookupCache.m.Store(hash, append(rts, tt))
  		return tt
  	}
  
  	// Look in known types for the same string representation.
  	str := funcStr(ft)
  	for _, tt := range typesByString(str) {
  		if haveIdenticalUnderlyingType(&ft.rtype, tt, true) {
  			return addToCache(tt)
  		}
  	}
  
  	// Populate the remaining fields of ft and store in cache.
  	ft.str = resolveReflectName(newName(str, "", false))
  	ft.ptrToThis = 0
  	return addToCache(&ft.rtype)
  }
  
  // funcStr builds a string representation of a funcType.
  func funcStr(ft *funcType) string {
  	repr := make([]byte, 0, 64)
  	repr = append(repr, "func("...)
  	for i, t := range ft.in() {
  		if i > 0 {
  			repr = append(repr, ", "...)
  		}
  		if ft.IsVariadic() && i == int(ft.inCount)-1 {
  			repr = append(repr, "..."...)
  			repr = append(repr, (*sliceType)(unsafe.Pointer(t)).elem.String()...)
  		} else {
  			repr = append(repr, t.String()...)
  		}
  	}
  	repr = append(repr, ')')
  	out := ft.out()
  	if len(out) == 1 {
  		repr = append(repr, ' ')
  	} else if len(out) > 1 {
  		repr = append(repr, " ("...)
  	}
  	for i, t := range out {
  		if i > 0 {
  			repr = append(repr, ", "...)
  		}
  		repr = append(repr, t.String()...)
  	}
  	if len(out) > 1 {
  		repr = append(repr, ')')
  	}
  	return string(repr)
  }
  
  // isReflexive reports whether the == operation on the type is reflexive.
  // That is, x == x for all values x of type t.
  func isReflexive(t *rtype) bool {
  	switch t.Kind() {
  	case Bool, Int, Int8, Int16, Int32, Int64, Uint, Uint8, Uint16, Uint32, Uint64, Uintptr, Chan, Ptr, String, UnsafePointer:
  		return true
  	case Float32, Float64, Complex64, Complex128, Interface:
  		return false
  	case Array:
  		tt := (*arrayType)(unsafe.Pointer(t))
  		return isReflexive(tt.elem)
  	case Struct:
  		tt := (*structType)(unsafe.Pointer(t))
  		for _, f := range tt.fields {
  			if !isReflexive(f.typ) {
  				return false
  			}
  		}
  		return true
  	default:
  		// Func, Map, Slice, Invalid
  		panic("isReflexive called on non-key type " + t.String())
  	}
  }
  
  // needKeyUpdate reports whether map overwrites require the key to be copied.
  func needKeyUpdate(t *rtype) bool {
  	switch t.Kind() {
  	case Bool, Int, Int8, Int16, Int32, Int64, Uint, Uint8, Uint16, Uint32, Uint64, Uintptr, Chan, Ptr, UnsafePointer:
  		return false
  	case Float32, Float64, Complex64, Complex128, Interface, String:
  		// Float keys can be updated from +0 to -0.
  		// String keys can be updated to use a smaller backing store.
  		// Interfaces might have floats of strings in them.
  		return true
  	case Array:
  		tt := (*arrayType)(unsafe.Pointer(t))
  		return needKeyUpdate(tt.elem)
  	case Struct:
  		tt := (*structType)(unsafe.Pointer(t))
  		for _, f := range tt.fields {
  			if needKeyUpdate(f.typ) {
  				return true
  			}
  		}
  		return false
  	default:
  		// Func, Map, Slice, Invalid
  		panic("needKeyUpdate called on non-key type " + t.String())
  	}
  }
  
  // Make sure these routines stay in sync with ../../runtime/map.go!
  // These types exist only for GC, so we only fill out GC relevant info.
  // Currently, that's just size and the GC program. We also fill in string
  // for possible debugging use.
  const (
  	bucketSize uintptr = 8
  	maxKeySize uintptr = 128
  	maxValSize uintptr = 128
  )
  
  func bucketOf(ktyp, etyp *rtype) *rtype {
  	// See comment on hmap.overflow in ../runtime/map.go.
  	var kind uint8
  	if ktyp.kind&kindNoPointers != 0 && etyp.kind&kindNoPointers != 0 &&
  		ktyp.size <= maxKeySize && etyp.size <= maxValSize {
  		kind = kindNoPointers
  	}
  
  	if ktyp.size > maxKeySize {
  		ktyp = PtrTo(ktyp).(*rtype)
  	}
  	if etyp.size > maxValSize {
  		etyp = PtrTo(etyp).(*rtype)
  	}
  
  	// Prepare GC data if any.
  	// A bucket is at most bucketSize*(1+maxKeySize+maxValSize)+2*ptrSize bytes,
  	// or 2072 bytes, or 259 pointer-size words, or 33 bytes of pointer bitmap.
  	// Note that since the key and value are known to be <= 128 bytes,
  	// they're guaranteed to have bitmaps instead of GC programs.
  	var gcdata *byte
  	var ptrdata uintptr
  	var overflowPad uintptr
  
  	// On NaCl, pad if needed to make overflow end at the proper struct alignment.
  	// On other systems, align > ptrSize is not possible.
  	if runtime.GOARCH == "amd64p32" && (ktyp.align > ptrSize || etyp.align > ptrSize) {
  		overflowPad = ptrSize
  	}
  	size := bucketSize*(1+ktyp.size+etyp.size) + overflowPad + ptrSize
  	if size&uintptr(ktyp.align-1) != 0 || size&uintptr(etyp.align-1) != 0 {
  		panic("reflect: bad size computation in MapOf")
  	}
  
  	if kind != kindNoPointers {
  		nptr := (bucketSize*(1+ktyp.size+etyp.size) + ptrSize) / ptrSize
  		mask := make([]byte, (nptr+7)/8)
  		base := bucketSize / ptrSize
  
  		if ktyp.kind&kindNoPointers == 0 {
  			if ktyp.kind&kindGCProg != 0 {
  				panic("reflect: unexpected GC program in MapOf")
  			}
  			kmask := (*[16]byte)(unsafe.Pointer(ktyp.gcdata))
  			for i := uintptr(0); i < ktyp.ptrdata/ptrSize; i++ {
  				if (kmask[i/8]>>(i%8))&1 != 0 {
  					for j := uintptr(0); j < bucketSize; j++ {
  						word := base + j*ktyp.size/ptrSize + i
  						mask[word/8] |= 1 << (word % 8)
  					}
  				}
  			}
  		}
  		base += bucketSize * ktyp.size / ptrSize
  
  		if etyp.kind&kindNoPointers == 0 {
  			if etyp.kind&kindGCProg != 0 {
  				panic("reflect: unexpected GC program in MapOf")
  			}
  			emask := (*[16]byte)(unsafe.Pointer(etyp.gcdata))
  			for i := uintptr(0); i < etyp.ptrdata/ptrSize; i++ {
  				if (emask[i/8]>>(i%8))&1 != 0 {
  					for j := uintptr(0); j < bucketSize; j++ {
  						word := base + j*etyp.size/ptrSize + i
  						mask[word/8] |= 1 << (word % 8)
  					}
  				}
  			}
  		}
  		base += bucketSize * etyp.size / ptrSize
  		base += overflowPad / ptrSize
  
  		word := base
  		mask[word/8] |= 1 << (word % 8)
  		gcdata = &mask[0]
  		ptrdata = (word + 1) * ptrSize
  
  		// overflow word must be last
  		if ptrdata != size {
  			panic("reflect: bad layout computation in MapOf")
  		}
  	}
  
  	b := &rtype{
  		align:   ptrSize,
  		size:    size,
  		kind:    kind,
  		ptrdata: ptrdata,
  		gcdata:  gcdata,
  	}
  	if overflowPad > 0 {
  		b.align = 8
  	}
  	s := "bucket(" + ktyp.String() + "," + etyp.String() + ")"
  	b.str = resolveReflectName(newName(s, "", false))
  	return b
  }
  
  // SliceOf returns the slice type with element type t.
  // For example, if t represents int, SliceOf(t) represents []int.
  func SliceOf(t Type) Type {
  	typ := t.(*rtype)
  
  	// Look in cache.
  	ckey := cacheKey{Slice, typ, nil, 0}
  	if slice, ok := lookupCache.Load(ckey); ok {
  		return slice.(Type)
  	}
  
  	// Look in known types.
  	s := "[]" + typ.String()
  	for _, tt := range typesByString(s) {
  		slice := (*sliceType)(unsafe.Pointer(tt))
  		if slice.elem == typ {
  			ti, _ := lookupCache.LoadOrStore(ckey, tt)
  			return ti.(Type)
  		}
  	}
  
  	// Make a slice type.
  	var islice interface{} = ([]unsafe.Pointer)(nil)
  	prototype := *(**sliceType)(unsafe.Pointer(&islice))
  	slice := *prototype
  	slice.tflag = 0
  	slice.str = resolveReflectName(newName(s, "", false))
  	slice.hash = fnv1(typ.hash, '[')
  	slice.elem = typ
  	slice.ptrToThis = 0
  
  	ti, _ := lookupCache.LoadOrStore(ckey, &slice.rtype)
  	return ti.(Type)
  }
  
  // The structLookupCache caches StructOf lookups.
  // StructOf does not share the common lookupCache since we need to pin
  // the memory associated with *structTypeFixedN.
  var structLookupCache struct {
  	sync.Mutex // Guards stores (but not loads) on m.
  
  	// m is a map[uint32][]Type keyed by the hash calculated in StructOf.
  	// Elements in m are append-only and thus safe for concurrent reading.
  	m sync.Map
  }
  
  type structTypeUncommon struct {
  	structType
  	u uncommonType
  }
  
  // A *rtype representing a struct is followed directly in memory by an
  // array of method objects representing the methods attached to the
  // struct. To get the same layout for a run time generated type, we
  // need an array directly following the uncommonType memory. The types
  // structTypeFixed4, ...structTypeFixedN are used to do this.
  //
  // A similar strategy is used for funcTypeFixed4, ...funcTypeFixedN.
  
  // TODO(crawshaw): as these structTypeFixedN and funcTypeFixedN structs
  // have no methods, they could be defined at runtime using the StructOf
  // function.
  
  type structTypeFixed4 struct {
  	structType
  	u uncommonType
  	m [4]method
  }
  
  type structTypeFixed8 struct {
  	structType
  	u uncommonType
  	m [8]method
  }
  
  type structTypeFixed16 struct {
  	structType
  	u uncommonType
  	m [16]method
  }
  
  type structTypeFixed32 struct {
  	structType
  	u uncommonType
  	m [32]method
  }
  
  // isLetter returns true if a given 'rune' is classified as a Letter.
  func isLetter(ch rune) bool {
  	return 'a' <= ch && ch <= 'z' || 'A' <= ch && ch <= 'Z' || ch == '_' || ch >= utf8.RuneSelf && unicode.IsLetter(ch)
  }
  
  // isValidFieldName checks if a string is a valid (struct) field name or not.
  //
  // According to the language spec, a field name should be an identifier.
  //
  // identifier = letter { letter | unicode_digit } .
  // letter = unicode_letter | "_" .
  func isValidFieldName(fieldName string) bool {
  	for i, c := range fieldName {
  		if i == 0 && !isLetter(c) {
  			return false
  		}
  
  		if !(isLetter(c) || unicode.IsDigit(c)) {
  			return false
  		}
  	}
  
  	return len(fieldName) > 0
  }
  
  // StructOf returns the struct type containing fields.
  // The Offset and Index fields are ignored and computed as they would be
  // by the compiler.
  //
  // StructOf currently does not generate wrapper methods for embedded
  // fields and panics if passed unexported StructFields.
  // These limitations may be lifted in a future version.
  func StructOf(fields []StructField) Type {
  	var (
  		hash       = fnv1(0, []byte("struct {")...)
  		size       uintptr
  		typalign   uint8
  		comparable = true
  		hashable   = true
  		methods    []method
  
  		fs   = make([]structField, len(fields))
  		repr = make([]byte, 0, 64)
  		fset = map[string]struct{}{} // fields' names
  
  		hasPtr    = false // records whether at least one struct-field is a pointer
  		hasGCProg = false // records whether a struct-field type has a GCProg
  	)
  
  	lastzero := uintptr(0)
  	repr = append(repr, "struct {"...)
  	for i, field := range fields {
  		if field.Name == "" {
  			panic("reflect.StructOf: field " + strconv.Itoa(i) + " has no name")
  		}
  		if !isValidFieldName(field.Name) {
  			panic("reflect.StructOf: field " + strconv.Itoa(i) + " has invalid name")
  		}
  		if field.Type == nil {
  			panic("reflect.StructOf: field " + strconv.Itoa(i) + " has no type")
  		}
  		f := runtimeStructField(field)
  		ft := f.typ
  		if ft.kind&kindGCProg != 0 {
  			hasGCProg = true
  		}
  		if ft.pointers() {
  			hasPtr = true
  		}
  
  		// Update string and hash
  		name := f.name.name()
  		hash = fnv1(hash, []byte(name)...)
  		repr = append(repr, (" " + name)...)
  		if f.embedded() {
  			// Embedded field
  			if f.typ.Kind() == Ptr {
  				// Embedded ** and *interface{} are illegal
  				elem := ft.Elem()
  				if k := elem.Kind(); k == Ptr || k == Interface {
  					panic("reflect.StructOf: illegal embedded field type " + ft.String())
  				}
  			}
  
  			switch f.typ.Kind() {
  			case Interface:
  				ift := (*interfaceType)(unsafe.Pointer(ft))
  				for im, m := range ift.methods {
  					if ift.nameOff(m.name).pkgPath() != "" {
  						// TODO(sbinet).  Issue 15924.
  						panic("reflect: embedded interface with unexported method(s) not implemented")
  					}
  
  					var (
  						mtyp    = ift.typeOff(m.typ)
  						ifield  = i
  						imethod = im
  						ifn     Value
  						tfn     Value
  					)
  
  					if ft.kind&kindDirectIface != 0 {
  						tfn = MakeFunc(mtyp, func(in []Value) []Value {
  							var args []Value
  							var recv = in[0]
  							if len(in) > 1 {
  								args = in[1:]
  							}
  							return recv.Field(ifield).Method(imethod).Call(args)
  						})
  						ifn = MakeFunc(mtyp, func(in []Value) []Value {
  							var args []Value
  							var recv = in[0]
  							if len(in) > 1 {
  								args = in[1:]
  							}
  							return recv.Field(ifield).Method(imethod).Call(args)
  						})
  					} else {
  						tfn = MakeFunc(mtyp, func(in []Value) []Value {
  							var args []Value
  							var recv = in[0]
  							if len(in) > 1 {
  								args = in[1:]
  							}
  							return recv.Field(ifield).Method(imethod).Call(args)
  						})
  						ifn = MakeFunc(mtyp, func(in []Value) []Value {
  							var args []Value
  							var recv = Indirect(in[0])
  							if len(in) > 1 {
  								args = in[1:]
  							}
  							return recv.Field(ifield).Method(imethod).Call(args)
  						})
  					}
  
  					methods = append(methods, method{
  						name: resolveReflectName(ift.nameOff(m.name)),
  						mtyp: resolveReflectType(mtyp),
  						ifn:  resolveReflectText(unsafe.Pointer(&ifn)),
  						tfn:  resolveReflectText(unsafe.Pointer(&tfn)),
  					})
  				}
  			case Ptr:
  				ptr := (*ptrType)(unsafe.Pointer(ft))
  				if unt := ptr.uncommon(); unt != nil {
  					if i > 0 && unt.mcount > 0 {
  						// Issue 15924.
  						panic("reflect: embedded type with methods not implemented if type is not first field")
  					}
  					if len(fields) > 1 {
  						panic("reflect: embedded type with methods not implemented if there is more than one field")
  					}
  					for _, m := range unt.methods() {
  						mname := ptr.nameOff(m.name)
  						if mname.pkgPath() != "" {
  							// TODO(sbinet).
  							// Issue 15924.
  							panic("reflect: embedded interface with unexported method(s) not implemented")
  						}
  						methods = append(methods, method{
  							name: resolveReflectName(mname),
  							mtyp: resolveReflectType(ptr.typeOff(m.mtyp)),
  							ifn:  resolveReflectText(ptr.textOff(m.ifn)),
  							tfn:  resolveReflectText(ptr.textOff(m.tfn)),
  						})
  					}
  				}
  				if unt := ptr.elem.uncommon(); unt != nil {
  					for _, m := range unt.methods() {
  						mname := ptr.nameOff(m.name)
  						if mname.pkgPath() != "" {
  							// TODO(sbinet)
  							// Issue 15924.
  							panic("reflect: embedded interface with unexported method(s) not implemented")
  						}
  						methods = append(methods, method{
  							name: resolveReflectName(mname),
  							mtyp: resolveReflectType(ptr.elem.typeOff(m.mtyp)),
  							ifn:  resolveReflectText(ptr.elem.textOff(m.ifn)),
  							tfn:  resolveReflectText(ptr.elem.textOff(m.tfn)),
  						})
  					}
  				}
  			default:
  				if unt := ft.uncommon(); unt != nil {
  					if i > 0 && unt.mcount > 0 {
  						// Issue 15924.
  						panic("reflect: embedded type with methods not implemented if type is not first field")
  					}
  					if len(fields) > 1 && ft.kind&kindDirectIface != 0 {
  						panic("reflect: embedded type with methods not implemented for non-pointer type")
  					}
  					for _, m := range unt.methods() {
  						mname := ft.nameOff(m.name)
  						if mname.pkgPath() != "" {
  							// TODO(sbinet)
  							// Issue 15924.
  							panic("reflect: embedded interface with unexported method(s) not implemented")
  						}
  						methods = append(methods, method{
  							name: resolveReflectName(mname),
  							mtyp: resolveReflectType(ft.typeOff(m.mtyp)),
  							ifn:  resolveReflectText(ft.textOff(m.ifn)),
  							tfn:  resolveReflectText(ft.textOff(m.tfn)),
  						})
  
  					}
  				}
  			}
  		}
  		if _, dup := fset[name]; dup {
  			panic("reflect.StructOf: duplicate field " + name)
  		}
  		fset[name] = struct{}{}
  
  		hash = fnv1(hash, byte(ft.hash>>24), byte(ft.hash>>16), byte(ft.hash>>8), byte(ft.hash))
  
  		repr = append(repr, (" " + ft.String())...)
  		if f.name.tagLen() > 0 {
  			hash = fnv1(hash, []byte(f.name.tag())...)
  			repr = append(repr, (" " + strconv.Quote(f.name.tag()))...)
  		}
  		if i < len(fields)-1 {
  			repr = append(repr, ';')
  		}
  
  		comparable = comparable && (ft.alg.equal != nil)
  		hashable = hashable && (ft.alg.hash != nil)
  
  		offset := align(size, uintptr(ft.align))
  		if ft.align > typalign {
  			typalign = ft.align
  		}
  		size = offset + ft.size
  		f.offsetEmbed |= offset << 1
  
  		if ft.size == 0 {
  			lastzero = size
  		}
  
  		fs[i] = f
  	}
  
  	if size > 0 && lastzero == size {
  		// This is a non-zero sized struct that ends in a
  		// zero-sized field. We add an extra byte of padding,
  		// to ensure that taking the address of the final
  		// zero-sized field can't manufacture a pointer to the
  		// next object in the heap. See issue 9401.
  		size++
  	}
  
  	var typ *structType
  	var ut *uncommonType
  
  	switch {
  	case len(methods) == 0:
  		t := new(structTypeUncommon)
  		typ = &t.structType
  		ut = &t.u
  	case len(methods) <= 4:
  		t := new(structTypeFixed4)
  		typ = &t.structType
  		ut = &t.u
  		copy(t.m[:], methods)
  	case len(methods) <= 8:
  		t := new(structTypeFixed8)
  		typ = &t.structType
  		ut = &t.u
  		copy(t.m[:], methods)
  	case len(methods) <= 16:
  		t := new(structTypeFixed16)
  		typ = &t.structType
  		ut = &t.u
  		copy(t.m[:], methods)
  	case len(methods) <= 32:
  		t := new(structTypeFixed32)
  		typ = &t.structType
  		ut = &t.u
  		copy(t.m[:], methods)
  	default:
  		panic("reflect.StructOf: too many methods")
  	}
  	// TODO(sbinet): Once we allow embedding multiple types,
  	// methods will need to be sorted like the compiler does.
  	// TODO(sbinet): Once we allow non-exported methods, we will
  	// need to compute xcount as the number of exported methods.
  	ut.mcount = uint16(len(methods))
  	ut.xcount = ut.mcount
  	ut.moff = uint32(unsafe.Sizeof(uncommonType{}))
  
  	if len(fs) > 0 {
  		repr = append(repr, ' ')
  	}
  	repr = append(repr, '}')
  	hash = fnv1(hash, '}')
  	str := string(repr)
  
  	// Round the size up to be a multiple of the alignment.
  	size = align(size, uintptr(typalign))
  
  	// Make the struct type.
  	var istruct interface{} = struct{}{}
  	prototype := *(**structType)(unsafe.Pointer(&istruct))
  	*typ = *prototype
  	typ.fields = fs
  
  	// Look in cache.
  	if ts, ok := structLookupCache.m.Load(hash); ok {
  		for _, st := range ts.([]Type) {
  			t := st.common()
  			if haveIdenticalUnderlyingType(&typ.rtype, t, true) {
  				return t
  			}
  		}
  	}
  
  	// Not in cache, lock and retry.
  	structLookupCache.Lock()
  	defer structLookupCache.Unlock()
  	if ts, ok := structLookupCache.m.Load(hash); ok {
  		for _, st := range ts.([]Type) {
  			t := st.common()
  			if haveIdenticalUnderlyingType(&typ.rtype, t, true) {
  				return t
  			}
  		}
  	}
  
  	addToCache := func(t Type) Type {
  		var ts []Type
  		if ti, ok := structLookupCache.m.Load(hash); ok {
  			ts = ti.([]Type)
  		}
  		structLookupCache.m.Store(hash, append(ts, t))
  		return t
  	}
  
  	// Look in known types.
  	for _, t := range typesByString(str) {
  		if haveIdenticalUnderlyingType(&typ.rtype, t, true) {
  			// even if 't' wasn't a structType with methods, we should be ok
  			// as the 'u uncommonType' field won't be accessed except when
  			// tflag&tflagUncommon is set.
  			return addToCache(t)
  		}
  	}
  
  	typ.str = resolveReflectName(newName(str, "", false))
  	typ.tflag = 0
  	typ.hash = hash
  	typ.size = size
  	typ.align = typalign
  	typ.fieldAlign = typalign
  	typ.ptrToThis = 0
  	if len(methods) > 0 {
  		typ.tflag |= tflagUncommon
  	}
  	if !hasPtr {
  		typ.kind |= kindNoPointers
  	} else {
  		typ.kind &^= kindNoPointers
  	}
  
  	if hasGCProg {
  		lastPtrField := 0
  		for i, ft := range fs {
  			if ft.typ.pointers() {
  				lastPtrField = i
  			}
  		}
  		prog := []byte{0, 0, 0, 0} // will be length of prog
  		for i, ft := range fs {
  			if i > lastPtrField {
  				// gcprog should not include anything for any field after
  				// the last field that contains pointer data
  				break
  			}
  			// FIXME(sbinet) handle padding, fields smaller than a word
  			elemGC := (*[1 << 30]byte)(unsafe.Pointer(ft.typ.gcdata))[:]
  			elemPtrs := ft.typ.ptrdata / ptrSize
  			switch {
  			case ft.typ.kind&kindGCProg == 0 && ft.typ.ptrdata != 0:
  				// Element is small with pointer mask; use as literal bits.
  				mask := elemGC
  				// Emit 120-bit chunks of full bytes (max is 127 but we avoid using partial bytes).
  				var n uintptr
  				for n := elemPtrs; n > 120; n -= 120 {
  					prog = append(prog, 120)
  					prog = append(prog, mask[:15]...)
  					mask = mask[15:]
  				}
  				prog = append(prog, byte(n))
  				prog = append(prog, mask[:(n+7)/8]...)
  			case ft.typ.kind&kindGCProg != 0:
  				// Element has GC program; emit one element.
  				elemProg := elemGC[4 : 4+*(*uint32)(unsafe.Pointer(&elemGC[0]))-1]
  				prog = append(prog, elemProg...)
  			}
  			// Pad from ptrdata to size.
  			elemWords := ft.typ.size / ptrSize
  			if elemPtrs < elemWords {
  				// Emit literal 0 bit, then repeat as needed.
  				prog = append(prog, 0x01, 0x00)
  				if elemPtrs+1 < elemWords {
  					prog = append(prog, 0x81)
  					prog = appendVarint(prog, elemWords-elemPtrs-1)
  				}
  			}
  		}
  		*(*uint32)(unsafe.Pointer(&prog[0])) = uint32(len(prog) - 4)
  		typ.kind |= kindGCProg
  		typ.gcdata = &prog[0]
  	} else {
  		typ.kind &^= kindGCProg
  		bv := new(bitVector)
  		addTypeBits(bv, 0, typ.common())
  		if len(bv.data) > 0 {
  			typ.gcdata = &bv.data[0]
  		}
  	}
  	typ.ptrdata = typeptrdata(typ.common())
  	typ.alg = new(typeAlg)
  	if hashable {
  		typ.alg.hash = func(p unsafe.Pointer, seed uintptr) uintptr {
  			o := seed
  			for _, ft := range typ.fields {
  				pi := add(p, ft.offset(), "&x.field safe")
  				o = ft.typ.alg.hash(pi, o)
  			}
  			return o
  		}
  	}
  
  	if comparable {
  		typ.alg.equal = func(p, q unsafe.Pointer) bool {
  			for _, ft := range typ.fields {
  				pi := add(p, ft.offset(), "&x.field safe")
  				qi := add(q, ft.offset(), "&x.field safe")
  				if !ft.typ.alg.equal(pi, qi) {
  					return false
  				}
  			}
  			return true
  		}
  	}
  
  	switch {
  	case len(fs) == 1 && !ifaceIndir(fs[0].typ):
  		// structs of 1 direct iface type can be direct
  		typ.kind |= kindDirectIface
  	default:
  		typ.kind &^= kindDirectIface
  	}
  
  	return addToCache(&typ.rtype)
  }
  
  func runtimeStructField(field StructField) structField {
  	if field.PkgPath != "" {
  		panic("reflect.StructOf: StructOf does not allow unexported fields")
  	}
  
  	// Best-effort check for misuse.
  	// Since PkgPath is empty, not much harm done if Unicode lowercase slips through.
  	c := field.Name[0]
  	if 'a' <= c && c <= 'z' || c == '_' {
  		panic("reflect.StructOf: field \"" + field.Name + "\" is unexported but missing PkgPath")
  	}
  
  	offsetEmbed := uintptr(0)
  	if field.Anonymous {
  		offsetEmbed |= 1
  	}
  
  	resolveReflectType(field.Type.common()) // install in runtime
  	return structField{
  		name:        newName(field.Name, string(field.Tag), true),
  		typ:         field.Type.common(),
  		offsetEmbed: offsetEmbed,
  	}
  }
  
  // typeptrdata returns the length in bytes of the prefix of t
  // containing pointer data. Anything after this offset is scalar data.
  // keep in sync with ../cmd/compile/internal/gc/reflect.go
  func typeptrdata(t *rtype) uintptr {
  	if !t.pointers() {
  		return 0
  	}
  	switch t.Kind() {
  	case Struct:
  		st := (*structType)(unsafe.Pointer(t))
  		// find the last field that has pointers.
  		field := 0
  		for i := range st.fields {
  			ft := st.fields[i].typ
  			if ft.pointers() {
  				field = i
  			}
  		}
  		f := st.fields[field]
  		return f.offset() + f.typ.ptrdata
  
  	default:
  		panic("reflect.typeptrdata: unexpected type, " + t.String())
  	}
  }
  
  // See cmd/compile/internal/gc/reflect.go for derivation of constant.
  const maxPtrmaskBytes = 2048
  
  // ArrayOf returns the array type with the given count and element type.
  // For example, if t represents int, ArrayOf(5, t) represents [5]int.
  //
  // If the resulting type would be larger than the available address space,
  // ArrayOf panics.
  func ArrayOf(count int, elem Type) Type {
  	typ := elem.(*rtype)
  
  	// Look in cache.
  	ckey := cacheKey{Array, typ, nil, uintptr(count)}
  	if array, ok := lookupCache.Load(ckey); ok {
  		return array.(Type)
  	}
  
  	// Look in known types.
  	s := "[" + strconv.Itoa(count) + "]" + typ.String()
  	for _, tt := range typesByString(s) {
  		array := (*arrayType)(unsafe.Pointer(tt))
  		if array.elem == typ {
  			ti, _ := lookupCache.LoadOrStore(ckey, tt)
  			return ti.(Type)
  		}
  	}
  
  	// Make an array type.
  	var iarray interface{} = [1]unsafe.Pointer{}
  	prototype := *(**arrayType)(unsafe.Pointer(&iarray))
  	array := *prototype
  	array.tflag = 0
  	array.str = resolveReflectName(newName(s, "", false))
  	array.hash = fnv1(typ.hash, '[')
  	for n := uint32(count); n > 0; n >>= 8 {
  		array.hash = fnv1(array.hash, byte(n))
  	}
  	array.hash = fnv1(array.hash, ']')
  	array.elem = typ
  	array.ptrToThis = 0
  	if typ.size > 0 {
  		max := ^uintptr(0) / typ.size
  		if uintptr(count) > max {
  			panic("reflect.ArrayOf: array size would exceed virtual address space")
  		}
  	}
  	array.size = typ.size * uintptr(count)
  	if count > 0 && typ.ptrdata != 0 {
  		array.ptrdata = typ.size*uintptr(count-1) + typ.ptrdata
  	}
  	array.align = typ.align
  	array.fieldAlign = typ.fieldAlign
  	array.len = uintptr(count)
  	array.slice = SliceOf(elem).(*rtype)
  
  	array.kind &^= kindNoPointers
  	switch {
  	case typ.kind&kindNoPointers != 0 || array.size == 0:
  		// No pointers.
  		array.kind |= kindNoPointers
  		array.gcdata = nil
  		array.ptrdata = 0
  
  	case count == 1:
  		// In memory, 1-element array looks just like the element.
  		array.kind |= typ.kind & kindGCProg
  		array.gcdata = typ.gcdata
  		array.ptrdata = typ.ptrdata
  
  	case typ.kind&kindGCProg == 0 && array.size <= maxPtrmaskBytes*8*ptrSize:
  		// Element is small with pointer mask; array is still small.
  		// Create direct pointer mask by turning each 1 bit in elem
  		// into count 1 bits in larger mask.
  		mask := make([]byte, (array.ptrdata/ptrSize+7)/8)
  		elemMask := (*[1 << 30]byte)(unsafe.Pointer(typ.gcdata))[:]
  		elemWords := typ.size / ptrSize
  		for j := uintptr(0); j < typ.ptrdata/ptrSize; j++ {
  			if (elemMask[j/8]>>(j%8))&1 != 0 {
  				for i := uintptr(0); i < array.len; i++ {
  					k := i*elemWords + j
  					mask[k/8] |= 1 << (k % 8)
  				}
  			}
  		}
  		array.gcdata = &mask[0]
  
  	default:
  		// Create program that emits one element
  		// and then repeats to make the array.
  		prog := []byte{0, 0, 0, 0} // will be length of prog
  		elemGC := (*[1 << 30]byte)(unsafe.Pointer(typ.gcdata))[:]
  		elemPtrs := typ.ptrdata / ptrSize
  		if typ.kind&kindGCProg == 0 {
  			// Element is small with pointer mask; use as literal bits.
  			mask := elemGC
  			// Emit 120-bit chunks of full bytes (max is 127 but we avoid using partial bytes).
  			var n uintptr
  			for n = elemPtrs; n > 120; n -= 120 {
  				prog = append(prog, 120)
  				prog = append(prog, mask[:15]...)
  				mask = mask[15:]
  			}
  			prog = append(prog, byte(n))
  			prog = append(prog, mask[:(n+7)/8]...)
  		} else {
  			// Element has GC program; emit one element.
  			elemProg := elemGC[4 : 4+*(*uint32)(unsafe.Pointer(&elemGC[0]))-1]
  			prog = append(prog, elemProg...)
  		}
  		// Pad from ptrdata to size.
  		elemWords := typ.size / ptrSize
  		if elemPtrs < elemWords {
  			// Emit literal 0 bit, then repeat as needed.
  			prog = append(prog, 0x01, 0x00)
  			if elemPtrs+1 < elemWords {
  				prog = append(prog, 0x81)
  				prog = appendVarint(prog, elemWords-elemPtrs-1)
  			}
  		}
  		// Repeat count-1 times.
  		if elemWords < 0x80 {
  			prog = append(prog, byte(elemWords|0x80))
  		} else {
  			prog = append(prog, 0x80)
  			prog = appendVarint(prog, elemWords)
  		}
  		prog = appendVarint(prog, uintptr(count)-1)
  		prog = append(prog, 0)
  		*(*uint32)(unsafe.Pointer(&prog[0])) = uint32(len(prog) - 4)
  		array.kind |= kindGCProg
  		array.gcdata = &prog[0]
  		array.ptrdata = array.size // overestimate but ok; must match program
  	}
  
  	etyp := typ.common()
  	esize := etyp.Size()
  	ealg := etyp.alg
  
  	array.alg = new(typeAlg)
  	if ealg.equal != nil {
  		eequal := ealg.equal
  		array.alg.equal = func(p, q unsafe.Pointer) bool {
  			for i := 0; i < count; i++ {
  				pi := arrayAt(p, i, esize, "i < count")
  				qi := arrayAt(q, i, esize, "i < count")
  				if !eequal(pi, qi) {
  					return false
  				}
  
  			}
  			return true
  		}
  	}
  	if ealg.hash != nil {
  		ehash := ealg.hash
  		array.alg.hash = func(ptr unsafe.Pointer, seed uintptr) uintptr {
  			o := seed
  			for i := 0; i < count; i++ {
  				o = ehash(arrayAt(ptr, i, esize, "i < count"), o)
  			}
  			return o
  		}
  	}
  
  	switch {
  	case count == 1 && !ifaceIndir(typ):
  		// array of 1 direct iface type can be direct
  		array.kind |= kindDirectIface
  	default:
  		array.kind &^= kindDirectIface
  	}
  
  	ti, _ := lookupCache.LoadOrStore(ckey, &array.rtype)
  	return ti.(Type)
  }
  
  func appendVarint(x []byte, v uintptr) []byte {
  	for ; v >= 0x80; v >>= 7 {
  		x = append(x, byte(v|0x80))
  	}
  	x = append(x, byte(v))
  	return x
  }
  
  // toType converts from a *rtype to a Type that can be returned
  // to the client of package reflect. In gc, the only concern is that
  // a nil *rtype must be replaced by a nil Type, but in gccgo this
  // function takes care of ensuring that multiple *rtype for the same
  // type are coalesced into a single Type.
  func toType(t *rtype) Type {
  	if t == nil {
  		return nil
  	}
  	return t
  }
  
  type layoutKey struct {
  	t    *rtype // function signature
  	rcvr *rtype // receiver type, or nil if none
  }
  
  type layoutType struct {
  	t         *rtype
  	argSize   uintptr // size of arguments
  	retOffset uintptr // offset of return values.
  	stack     *bitVector
  	framePool *sync.Pool
  }
  
  var layoutCache sync.Map // map[layoutKey]layoutType
  
  // funcLayout computes a struct type representing the layout of the
  // function arguments and return values for the function type t.
  // If rcvr != nil, rcvr specifies the type of the receiver.
  // The returned type exists only for GC, so we only fill out GC relevant info.
  // Currently, that's just size and the GC program. We also fill in
  // the name for possible debugging use.
  func funcLayout(t *rtype, rcvr *rtype) (frametype *rtype, argSize, retOffset uintptr, stk *bitVector, framePool *sync.Pool) {
  	if t.Kind() != Func {
  		panic("reflect: funcLayout of non-func type")
  	}
  	if rcvr != nil && rcvr.Kind() == Interface {
  		panic("reflect: funcLayout with interface receiver " + rcvr.String())
  	}
  	k := layoutKey{t, rcvr}
  	if lti, ok := layoutCache.Load(k); ok {
  		lt := lti.(layoutType)
  		return lt.t, lt.argSize, lt.retOffset, lt.stack, lt.framePool
  	}
  
  	tt := (*funcType)(unsafe.Pointer(t))
  
  	// compute gc program & stack bitmap for arguments
  	ptrmap := new(bitVector)
  	var offset uintptr
  	if rcvr != nil {
  		// Reflect uses the "interface" calling convention for
  		// methods, where receivers take one word of argument
  		// space no matter how big they actually are.
  		if ifaceIndir(rcvr) || rcvr.pointers() {
  			ptrmap.append(1)
  		}
  		offset += ptrSize
  	}
  	for _, arg := range tt.in() {
  		offset += -offset & uintptr(arg.align-1)
  		addTypeBits(ptrmap, offset, arg)
  		offset += arg.size
  	}
  	argN := ptrmap.n
  	argSize = offset
  	if runtime.GOARCH == "amd64p32" {
  		offset += -offset & (8 - 1)
  	}
  	offset += -offset & (ptrSize - 1)
  	retOffset = offset
  	for _, res := range tt.out() {
  		offset += -offset & uintptr(res.align-1)
  		addTypeBits(ptrmap, offset, res)
  		offset += res.size
  	}
  	offset += -offset & (ptrSize - 1)
  
  	// build dummy rtype holding gc program
  	x := &rtype{
  		align:   ptrSize,
  		size:    offset,
  		ptrdata: uintptr(ptrmap.n) * ptrSize,
  	}
  	if runtime.GOARCH == "amd64p32" {
  		x.align = 8
  	}
  	if ptrmap.n > 0 {
  		x.gcdata = &ptrmap.data[0]
  	} else {
  		x.kind |= kindNoPointers
  	}
  	ptrmap.n = argN
  
  	var s string
  	if rcvr != nil {
  		s = "methodargs(" + rcvr.String() + ")(" + t.String() + ")"
  	} else {
  		s = "funcargs(" + t.String() + ")"
  	}
  	x.str = resolveReflectName(newName(s, "", false))
  
  	// cache result for future callers
  	framePool = &sync.Pool{New: func() interface{} {
  		return unsafe_New(x)
  	}}
  	lti, _ := layoutCache.LoadOrStore(k, layoutType{
  		t:         x,
  		argSize:   argSize,
  		retOffset: retOffset,
  		stack:     ptrmap,
  		framePool: framePool,
  	})
  	lt := lti.(layoutType)
  	return lt.t, lt.argSize, lt.retOffset, lt.stack, lt.framePool
  }
  
  // ifaceIndir reports whether t is stored indirectly in an interface value.
  func ifaceIndir(t *rtype) bool {
  	return t.kind&kindDirectIface == 0
  }
  
  // Layout matches runtime.gobitvector (well enough).
  type bitVector struct {
  	n    uint32 // number of bits
  	data []byte
  }
  
  // append a bit to the bitmap.
  func (bv *bitVector) append(bit uint8) {
  	if bv.n%8 == 0 {
  		bv.data = append(bv.data, 0)
  	}
  	bv.data[bv.n/8] |= bit << (bv.n % 8)
  	bv.n++
  }
  
  func addTypeBits(bv *bitVector, offset uintptr, t *rtype) {
  	if t.kind&kindNoPointers != 0 {
  		return
  	}
  
  	switch Kind(t.kind & kindMask) {
  	case Chan, Func, Map, Ptr, Slice, String, UnsafePointer:
  		// 1 pointer at start of representation
  		for bv.n < uint32(offset/uintptr(ptrSize)) {
  			bv.append(0)
  		}
  		bv.append(1)
  
  	case Interface:
  		// 2 pointers
  		for bv.n < uint32(offset/uintptr(ptrSize)) {
  			bv.append(0)
  		}
  		bv.append(1)
  		bv.append(1)
  
  	case Array:
  		// repeat inner type
  		tt := (*arrayType)(unsafe.Pointer(t))
  		for i := 0; i < int(tt.len); i++ {
  			addTypeBits(bv, offset+uintptr(i)*tt.elem.size, tt.elem)
  		}
  
  	case Struct:
  		// apply fields
  		tt := (*structType)(unsafe.Pointer(t))
  		for i := range tt.fields {
  			f := &tt.fields[i]
  			addTypeBits(bv, offset+f.offset(), f.typ)
  		}
  	}
  }
  

View as plain text