go noder 源码

  • 2022-07-15
  • 浏览 (922)

golang noder 代码

文件路径:/src/cmd/compile/internal/noder/noder.go

// Copyright 2016 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 noder

import (
	"errors"
	"fmt"
	"os"
	"path/filepath"
	"runtime"
	"strconv"
	"strings"
	"unicode"
	"unicode/utf8"

	"cmd/compile/internal/base"
	"cmd/compile/internal/ir"
	"cmd/compile/internal/syntax"
	"cmd/compile/internal/typecheck"
	"cmd/compile/internal/types"
	"cmd/internal/objabi"
	"cmd/internal/src"
)

func LoadPackage(filenames []string) {
	base.Timer.Start("fe", "parse")

	// Limit the number of simultaneously open files.
	sem := make(chan struct{}, runtime.GOMAXPROCS(0)+10)

	noders := make([]*noder, len(filenames))
	for i := range noders {
		p := noder{
			err: make(chan syntax.Error),
		}
		noders[i] = &p
	}

	// Move the entire syntax processing logic into a separate goroutine to avoid blocking on the "sem".
	go func() {
		for i, filename := range filenames {
			filename := filename
			p := noders[i]
			sem <- struct{}{}
			go func() {
				defer func() { <-sem }()
				defer close(p.err)
				fbase := syntax.NewFileBase(filename)

				f, err := os.Open(filename)
				if err != nil {
					p.error(syntax.Error{Msg: err.Error()})
					return
				}
				defer f.Close()

				p.file, _ = syntax.Parse(fbase, f, p.error, p.pragma, syntax.CheckBranches) // errors are tracked via p.error
			}()
		}
	}()

	var lines uint
	for _, p := range noders {
		for e := range p.err {
			p.errorAt(e.Pos, "%s", e.Msg)
		}
		if p.file == nil {
			base.ErrorExit()
		}
		lines += p.file.EOF.Line()
	}
	base.Timer.AddEvent(int64(lines), "lines")

	if base.Debug.Unified != 0 {
		unified(noders)
		return
	}

	// Use types2 to type-check and generate IR.
	check2(noders)
}

func (p *noder) errorAt(pos syntax.Pos, format string, args ...interface{}) {
	base.ErrorfAt(p.makeXPos(pos), format, args...)
}

// trimFilename returns the "trimmed" filename of b, which is the
// absolute filename after applying -trimpath processing. This
// filename form is suitable for use in object files and export data.
//
// If b's filename has already been trimmed (i.e., because it was read
// in from an imported package's export data), then the filename is
// returned unchanged.
func trimFilename(b *syntax.PosBase) string {
	filename := b.Filename()
	if !b.Trimmed() {
		dir := ""
		if b.IsFileBase() {
			dir = base.Ctxt.Pathname
		}
		filename = objabi.AbsFile(dir, filename, base.Flag.TrimPath)
	}
	return filename
}

// noder transforms package syntax's AST into a Node tree.
type noder struct {
	posMap

	file           *syntax.File
	linknames      []linkname
	pragcgobuf     [][]string
	err            chan syntax.Error
	importedUnsafe bool
	importedEmbed  bool
}

// linkname records a //go:linkname directive.
type linkname struct {
	pos    syntax.Pos
	local  string
	remote string
}

func (p *noder) processPragmas() {
	for _, l := range p.linknames {
		if !p.importedUnsafe {
			p.errorAt(l.pos, "//go:linkname only allowed in Go files that import \"unsafe\"")
			continue
		}
		n := ir.AsNode(typecheck.Lookup(l.local).Def)
		if n == nil || n.Op() != ir.ONAME {
			p.errorAt(l.pos, "//go:linkname must refer to declared function or variable")
			continue
		}
		if n.Sym().Linkname != "" {
			p.errorAt(l.pos, "duplicate //go:linkname for %s", l.local)
			continue
		}
		n.Sym().Linkname = l.remote
	}
	typecheck.Target.CgoPragmas = append(typecheck.Target.CgoPragmas, p.pragcgobuf...)
}

var unOps = [...]ir.Op{
	syntax.Recv: ir.ORECV,
	syntax.Mul:  ir.ODEREF,
	syntax.And:  ir.OADDR,

	syntax.Not: ir.ONOT,
	syntax.Xor: ir.OBITNOT,
	syntax.Add: ir.OPLUS,
	syntax.Sub: ir.ONEG,
}

var binOps = [...]ir.Op{
	syntax.OrOr:   ir.OOROR,
	syntax.AndAnd: ir.OANDAND,

	syntax.Eql: ir.OEQ,
	syntax.Neq: ir.ONE,
	syntax.Lss: ir.OLT,
	syntax.Leq: ir.OLE,
	syntax.Gtr: ir.OGT,
	syntax.Geq: ir.OGE,

	syntax.Add: ir.OADD,
	syntax.Sub: ir.OSUB,
	syntax.Or:  ir.OOR,
	syntax.Xor: ir.OXOR,

	syntax.Mul:    ir.OMUL,
	syntax.Div:    ir.ODIV,
	syntax.Rem:    ir.OMOD,
	syntax.And:    ir.OAND,
	syntax.AndNot: ir.OANDNOT,
	syntax.Shl:    ir.OLSH,
	syntax.Shr:    ir.ORSH,
}

func wrapname(pos src.XPos, x ir.Node) ir.Node {
	// These nodes do not carry line numbers.
	// Introduce a wrapper node to give them the correct line.
	switch x.Op() {
	case ir.OTYPE, ir.OLITERAL:
		if x.Sym() == nil {
			break
		}
		fallthrough
	case ir.ONAME, ir.ONONAME:
		p := ir.NewParenExpr(pos, x)
		p.SetImplicit(true)
		return p
	}
	return x
}

// error is called concurrently if files are parsed concurrently.
func (p *noder) error(err error) {
	p.err <- err.(syntax.Error)
}

// pragmas that are allowed in the std lib, but don't have
// a syntax.Pragma value (see lex.go) associated with them.
var allowedStdPragmas = map[string]bool{
	"go:cgo_export_static":  true,
	"go:cgo_export_dynamic": true,
	"go:cgo_import_static":  true,
	"go:cgo_import_dynamic": true,
	"go:cgo_ldflag":         true,
	"go:cgo_dynamic_linker": true,
	"go:embed":              true,
	"go:generate":           true,
}

// *pragmas is the value stored in a syntax.pragmas during parsing.
type pragmas struct {
	Flag   ir.PragmaFlag // collected bits
	Pos    []pragmaPos   // position of each individual flag
	Embeds []pragmaEmbed
}

type pragmaPos struct {
	Flag ir.PragmaFlag
	Pos  syntax.Pos
}

type pragmaEmbed struct {
	Pos      syntax.Pos
	Patterns []string
}

func (p *noder) checkUnusedDuringParse(pragma *pragmas) {
	for _, pos := range pragma.Pos {
		if pos.Flag&pragma.Flag != 0 {
			p.error(syntax.Error{Pos: pos.Pos, Msg: "misplaced compiler directive"})
		}
	}
	if len(pragma.Embeds) > 0 {
		for _, e := range pragma.Embeds {
			p.error(syntax.Error{Pos: e.Pos, Msg: "misplaced go:embed directive"})
		}
	}
}

// pragma is called concurrently if files are parsed concurrently.
func (p *noder) pragma(pos syntax.Pos, blankLine bool, text string, old syntax.Pragma) syntax.Pragma {
	pragma, _ := old.(*pragmas)
	if pragma == nil {
		pragma = new(pragmas)
	}

	if text == "" {
		// unused pragma; only called with old != nil.
		p.checkUnusedDuringParse(pragma)
		return nil
	}

	if strings.HasPrefix(text, "line ") {
		// line directives are handled by syntax package
		panic("unreachable")
	}

	if !blankLine {
		// directive must be on line by itself
		p.error(syntax.Error{Pos: pos, Msg: "misplaced compiler directive"})
		return pragma
	}

	switch {
	case strings.HasPrefix(text, "go:linkname "):
		f := strings.Fields(text)
		if !(2 <= len(f) && len(f) <= 3) {
			p.error(syntax.Error{Pos: pos, Msg: "usage: //go:linkname localname [linkname]"})
			break
		}
		// The second argument is optional. If omitted, we use
		// the default object symbol name for this and
		// linkname only serves to mark this symbol as
		// something that may be referenced via the object
		// symbol name from another package.
		var target string
		if len(f) == 3 {
			target = f[2]
		} else if base.Ctxt.Pkgpath != "" {
			// Use the default object symbol name if the
			// user didn't provide one.
			target = objabi.PathToPrefix(base.Ctxt.Pkgpath) + "." + f[1]
		} else {
			p.error(syntax.Error{Pos: pos, Msg: "//go:linkname requires linkname argument or -p compiler flag"})
			break
		}
		p.linknames = append(p.linknames, linkname{pos, f[1], target})

	case text == "go:embed", strings.HasPrefix(text, "go:embed "):
		args, err := parseGoEmbed(text[len("go:embed"):])
		if err != nil {
			p.error(syntax.Error{Pos: pos, Msg: err.Error()})
		}
		if len(args) == 0 {
			p.error(syntax.Error{Pos: pos, Msg: "usage: //go:embed pattern..."})
			break
		}
		pragma.Embeds = append(pragma.Embeds, pragmaEmbed{pos, args})

	case strings.HasPrefix(text, "go:cgo_import_dynamic "):
		// This is permitted for general use because Solaris
		// code relies on it in golang.org/x/sys/unix and others.
		fields := pragmaFields(text)
		if len(fields) >= 4 {
			lib := strings.Trim(fields[3], `"`)
			if lib != "" && !safeArg(lib) && !isCgoGeneratedFile(pos) {
				p.error(syntax.Error{Pos: pos, Msg: fmt.Sprintf("invalid library name %q in cgo_import_dynamic directive", lib)})
			}
			p.pragcgo(pos, text)
			pragma.Flag |= pragmaFlag("go:cgo_import_dynamic")
			break
		}
		fallthrough
	case strings.HasPrefix(text, "go:cgo_"):
		// For security, we disallow //go:cgo_* directives other
		// than cgo_import_dynamic outside cgo-generated files.
		// Exception: they are allowed in the standard library, for runtime and syscall.
		if !isCgoGeneratedFile(pos) && !base.Flag.Std {
			p.error(syntax.Error{Pos: pos, Msg: fmt.Sprintf("//%s only allowed in cgo-generated code", text)})
		}
		p.pragcgo(pos, text)
		fallthrough // because of //go:cgo_unsafe_args
	default:
		verb := text
		if i := strings.Index(text, " "); i >= 0 {
			verb = verb[:i]
		}
		flag := pragmaFlag(verb)
		const runtimePragmas = ir.Systemstack | ir.Nowritebarrier | ir.Nowritebarrierrec | ir.Yeswritebarrierrec
		if !base.Flag.CompilingRuntime && flag&runtimePragmas != 0 {
			p.error(syntax.Error{Pos: pos, Msg: fmt.Sprintf("//%s only allowed in runtime", verb)})
		}
		if flag == ir.UintptrKeepAlive && !base.Flag.Std {
			p.error(syntax.Error{Pos: pos, Msg: fmt.Sprintf("//%s is only allowed in the standard library", verb)})
		}
		if flag == 0 && !allowedStdPragmas[verb] && base.Flag.Std {
			p.error(syntax.Error{Pos: pos, Msg: fmt.Sprintf("//%s is not allowed in the standard library", verb)})
		}
		pragma.Flag |= flag
		pragma.Pos = append(pragma.Pos, pragmaPos{flag, pos})
	}

	return pragma
}

// isCgoGeneratedFile reports whether pos is in a file
// generated by cgo, which is to say a file with name
// beginning with "_cgo_". Such files are allowed to
// contain cgo directives, and for security reasons
// (primarily misuse of linker flags), other files are not.
// See golang.org/issue/23672.
func isCgoGeneratedFile(pos syntax.Pos) bool {
	return strings.HasPrefix(filepath.Base(trimFilename(pos.Base())), "_cgo_")
}

// safeArg reports whether arg is a "safe" command-line argument,
// meaning that when it appears in a command-line, it probably
// doesn't have some special meaning other than its own name.
// This is copied from SafeArg in cmd/go/internal/load/pkg.go.
func safeArg(name string) bool {
	if name == "" {
		return false
	}
	c := name[0]
	return '0' <= c && c <= '9' || 'A' <= c && c <= 'Z' || 'a' <= c && c <= 'z' || c == '.' || c == '_' || c == '/' || c >= utf8.RuneSelf
}

// parseGoEmbed parses the text following "//go:embed" to extract the glob patterns.
// It accepts unquoted space-separated patterns as well as double-quoted and back-quoted Go strings.
// go/build/read.go also processes these strings and contains similar logic.
func parseGoEmbed(args string) ([]string, error) {
	var list []string
	for args = strings.TrimSpace(args); args != ""; args = strings.TrimSpace(args) {
		var path string
	Switch:
		switch args[0] {
		default:
			i := len(args)
			for j, c := range args {
				if unicode.IsSpace(c) {
					i = j
					break
				}
			}
			path = args[:i]
			args = args[i:]

		case '`':
			i := strings.Index(args[1:], "`")
			if i < 0 {
				return nil, fmt.Errorf("invalid quoted string in //go:embed: %s", args)
			}
			path = args[1 : 1+i]
			args = args[1+i+1:]

		case '"':
			i := 1
			for ; i < len(args); i++ {
				if args[i] == '\\' {
					i++
					continue
				}
				if args[i] == '"' {
					q, err := strconv.Unquote(args[:i+1])
					if err != nil {
						return nil, fmt.Errorf("invalid quoted string in //go:embed: %s", args[:i+1])
					}
					path = q
					args = args[i+1:]
					break Switch
				}
			}
			if i >= len(args) {
				return nil, fmt.Errorf("invalid quoted string in //go:embed: %s", args)
			}
		}

		if args != "" {
			r, _ := utf8.DecodeRuneInString(args)
			if !unicode.IsSpace(r) {
				return nil, fmt.Errorf("invalid quoted string in //go:embed: %s", args)
			}
		}
		list = append(list, path)
	}
	return list, nil
}

// A function named init is a special case.
// It is called by the initialization before main is run.
// To make it unique within a package and also uncallable,
// the name, normally "pkg.init", is altered to "pkg.init.0".
var renameinitgen int

func Renameinit() *types.Sym {
	s := typecheck.LookupNum("init.", renameinitgen)
	renameinitgen++
	return s
}

func varEmbed(makeXPos func(syntax.Pos) src.XPos, name *ir.Name, decl *syntax.VarDecl, pragma *pragmas, haveEmbed bool) {
	pragmaEmbeds := pragma.Embeds
	pragma.Embeds = nil
	if len(pragmaEmbeds) == 0 {
		return
	}

	if err := checkEmbed(decl, haveEmbed, typecheck.DeclContext != ir.PEXTERN); err != nil {
		base.ErrorfAt(makeXPos(pragmaEmbeds[0].Pos), "%s", err)
		return
	}

	var embeds []ir.Embed
	for _, e := range pragmaEmbeds {
		embeds = append(embeds, ir.Embed{Pos: makeXPos(e.Pos), Patterns: e.Patterns})
	}
	typecheck.Target.Embeds = append(typecheck.Target.Embeds, name)
	name.Embed = &embeds
}

func checkEmbed(decl *syntax.VarDecl, haveEmbed, withinFunc bool) error {
	switch {
	case !haveEmbed:
		return errors.New("go:embed only allowed in Go files that import \"embed\"")
	case len(decl.NameList) > 1:
		return errors.New("go:embed cannot apply to multiple vars")
	case decl.Values != nil:
		return errors.New("go:embed cannot apply to var with initializer")
	case decl.Type == nil:
		// Should not happen, since Values == nil now.
		return errors.New("go:embed cannot apply to var without type")
	case withinFunc:
		return errors.New("go:embed cannot apply to var inside func")
	case !types.AllowsGoVersion(1, 16):
		return fmt.Errorf("go:embed requires go1.16 or later (-lang was set to %s; check go.mod)", base.Flag.Lang)

	default:
		return nil
	}
}

相关信息

go 源码目录

相关文章

go codes 源码

go decl 源码

go export 源码

go expr 源码

go func 源码

go helpers 源码

go import 源码

go irgen 源码

go lex 源码

go lex_test 源码

0  赞