go/parser: better error synchronization

gofmt reports now a single, accurate error for
the test case of issue 3106.

Also: Added test harness for general error
checking and two test cases for now.

Fixes #3106.

R=rsc, bradfitz
CC=golang-dev
https://golang.org/cl/5755062

src/pkg/go/parser/error_test.go

@@ -0,0 +1,163 @@
+// Copyright 2012 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.
+
+// This file implements a parser test harness. The files in the testdata
+// directory are parsed and the errors reported are compared against the
+// error messages expected in the test files. The test files must end in
+// .src rather than .go so that they are not disturbed by gofmt runs.
+//
+// Expected errors are indicated in the test files by putting a comment
+// of the form /* ERROR "rx" */ immediately following an offending token.
+// The harness will verify that an error matching the regular expression
+// rx is reported at that source position.
+//
+// For instance, the following test file indicates that a "not declared"
+// error should be reported for the undeclared variable x:
+//
+//	package p
+//	func f() {
+//		_ = x /* ERROR "not declared" */ + 1
+//	}
+
+package parser
+
+import (
+	"go/scanner"
+	"go/token"
+	"io/ioutil"
+	"path/filepath"
+	"regexp"
+	"strings"
+	"testing"
+)
+
+const testdata = "testdata"
+
+func getFile(filename string) (file *token.File) {
+	fset.Iterate(func(f *token.File) bool {
+		if f.Name() == filename {
+			file = f
+			return false // end iteration
+		}
+		return true
+	})
+	return file
+}
+
+func getPos(filename string, offset int) token.Pos {
+	if f := getFile(filename); f != nil {
+		return f.Pos(offset)
+	}
+	return token.NoPos
+}
+
+// ERROR comments must be of the form /* ERROR "rx" */ and rx is
+// a regular expression that matches the expected error message.
+//
+var errRx = regexp.MustCompile(`^/\* *ERROR *"([^"]*)" *\*/$`)
+
+// expectedErrors collects the regular expressions of ERROR comments found
+// in files and returns them as a map of error positions to error messages.
+//
+func expectedErrors(t *testing.T, filename string, src []byte) map[token.Pos]string {
+	errors := make(map[token.Pos]string)
+
+	var s scanner.Scanner
+	// file was parsed already - do not add it again to the file
+	// set otherwise the position information returned here will
+	// not match the position information collected by the parser
+	s.Init(getFile(filename), src, nil, scanner.ScanComments)
+	var prev token.Pos // position of last non-comment, non-semicolon token
+
+	for {
+		pos, tok, lit := s.Scan()
+		switch tok {
+		case token.EOF:
+			return errors
+		case token.COMMENT:
+			s := errRx.FindStringSubmatch(lit)
+			if len(s) == 2 {
+				errors[prev] = string(s[1])
+			}
+		default:
+			prev = pos
+		}
+	}
+
+	panic("unreachable")
+}
+
+// compareErrors compares the map of expected error messages with the list
+// of found errors and reports discrepancies.
+//
+func compareErrors(t *testing.T, expected map[token.Pos]string, found scanner.ErrorList) {
+	for _, error := range found {
+		// error.Pos is a token.Position, but we want
+		// a token.Pos so we can do a map lookup
+		pos := getPos(error.Pos.Filename, error.Pos.Offset)
+		if msg, found := expected[pos]; found {
+			// we expect a message at pos; check if it matches
+			rx, err := regexp.Compile(msg)
+			if err != nil {
+				t.Errorf("%s: %v", error.Pos, err)
+				continue
+			}
+			if match := rx.MatchString(error.Msg); !match {
+				t.Errorf("%s: %q does not match %q", error.Pos, error.Msg, msg)
+				continue
+			}
+			// we have a match - eliminate this error
+			delete(expected, pos)
+		} else {
+			// To keep in mind when analyzing failed test output:
+			// If the same error position occurs multiple times in errors,
+			// this message will be triggered (because the first error at
+			// the position removes this position from the expected errors).
+			t.Errorf("%s: unexpected error: %s", error.Pos, error.Msg)
+		}
+	}
+
+	// there should be no expected errors left
+	if len(expected) > 0 {
+		t.Errorf("%d errors not reported:", len(expected))
+		for pos, msg := range expected {
+			t.Errorf("%s: %s\n", fset.Position(pos), msg)
+		}
+	}
+}
+
+func checkErrors(t *testing.T, filename string) {
+	src, err := ioutil.ReadFile(filename)
+	if err != nil {
+		t.Error(err)
+		return
+	}
+
+	_, err = ParseFile(fset, filename, src, DeclarationErrors)
+	found, ok := err.(scanner.ErrorList)
+	if err != nil && !ok {
+		t.Error(err)
+		return
+	}
+
+	// we are expecting the following errors
+	// (collect these after parsing a file so that it is found in the file set)
+	expected := expectedErrors(t, filename, src)
+
+	// verify errors returned by the parser
+	compareErrors(t, expected, found)
+}
+
+func TestErrors(t *testing.T) {
+	list, err := ioutil.ReadDir(testdata)
+	if err != nil {
+		t.Fatal(err)
+	}
+	for _, fi := range list {
+		name := fi.Name()
+		if !fi.IsDir() && !strings.HasPrefix(name, ".") && strings.HasSuffix(name, ".src") {
+			checkErrors(t, filepath.Join(testdata, name))
+		}
+	}
+}

src/pkg/go/parser/parser.go

@@ -371,8 +371,16 @@ func (p *parser) expectClosing(tok token.Token, context string) token.Pos {
 }
 
 func (p *parser) expectSemi() {
+	// semicolon is optional before a closing ')' or '}'
 	if p.tok != token.RPAREN && p.tok != token.RBRACE {
-		p.expect(token.SEMICOLON)
+		if p.tok == token.SEMICOLON {
+			p.next()
+		} else {
+			p.errorExpected(p.pos, "';'")
+			for !isStmtSync(p.tok) {
+				p.next() // make progress
+			}
+		}
 	}
 }
 
@@ -394,6 +402,31 @@ func assert(cond bool, msg string) {
 	}
 }
 
+// isStmtSync reports whether tok starts a new statement.
+// Used for synchronization after an error.
+//
+func isStmtSync(tok token.Token) bool {
+	switch tok {
+	case token.BREAK, token.CONST, token.CONTINUE, token.DEFER,
+		token.FALLTHROUGH, token.FOR, token.GO, token.GOTO,
+		token.IF, token.RETURN, token.SELECT, token.SWITCH,
+		token.TYPE, token.VAR, token.EOF:
+		return true
+	}
+	return false
+}
+
+// isDeclSync reports whether tok starts a new declaration.
+// Used for synchronization after an error.
+//
+func isDeclSync(tok token.Token) bool {
+	switch tok {
+	case token.CONST, token.TYPE, token.VAR, token.EOF:
+		return true
+	}
+	return false
+}
+
 // ----------------------------------------------------------------------------
 // Identifiers
 
@@ -534,9 +567,11 @@ func (p *parser) makeIdentList(list []ast.Expr) []*ast.Ident {
 	for i, x := range list {
 		ident, isIdent := x.(*ast.Ident)
 		if !isIdent {
-			pos := x.Pos()
-			p.errorExpected(pos, "identifier")
-			ident = &ast.Ident{NamePos: pos, Name: "_"}
+			if _, isBad := x.(*ast.BadExpr); !isBad {
+				// only report error if it's a new one
+				p.errorExpected(x.Pos(), "identifier")
+			}
+			ident = &ast.Ident{NamePos: x.Pos(), Name: "_"}
 		}
 		idents[i] = ident
 	}
@@ -1003,19 +1038,21 @@ func (p *parser) parseOperand(lhs bool) ast.Expr {
 
 	case token.FUNC:
 		return p.parseFuncTypeOrLit()
-
-	default:
-		if typ := p.tryIdentOrType(true); typ != nil {
-			// could be type for composite literal or conversion
-			_, isIdent := typ.(*ast.Ident)
-			assert(!isIdent, "type cannot be identifier")
-			return typ
-		}
 	}
 
+	if typ := p.tryIdentOrType(true); typ != nil {
+		// could be type for composite literal or conversion
+		_, isIdent := typ.(*ast.Ident)
+		assert(!isIdent, "type cannot be identifier")
+		return typ
+	}
+
+	// we have an error
 	pos := p.pos
 	p.errorExpected(pos, "operand")
-	p.next() // make progress
+	if !isStmtSync(p.tok) {
+		p.next() // make progress
+	}
 	return &ast.BadExpr{From: pos, To: p.pos}
 }
 
@@ -1274,8 +1311,8 @@ L:
 				x = p.parseTypeAssertion(p.checkExpr(x))
 			default:
 				pos := p.pos
-				p.next() // make progress
 				p.errorExpected(pos, "selector or type assertion")
+				p.next() // make progress
 				x = &ast.BadExpr{From: pos, To: p.pos}
 			}
 		case token.LBRACK:
@@ -1483,7 +1520,10 @@ func (p *parser) parseCallExpr() *ast.CallExpr {
 	if call, isCall := x.(*ast.CallExpr); isCall {
 		return call
 	}
-	p.errorExpected(x.Pos(), "function/method call")
+	if _, isBad := x.(*ast.BadExpr); !isBad {
+		// only report error if it's a new one
+		p.errorExpected(x.Pos(), "function/method call")
+	}
 	return nil
 }
 
@@ -1874,7 +1914,7 @@ func (p *parser) parseStmt() (s ast.Stmt) {
 
 	switch p.tok {
 	case token.CONST, token.TYPE, token.VAR:
-		s = &ast.DeclStmt{Decl: p.parseDecl()}
+		s = &ast.DeclStmt{Decl: p.parseDecl(isStmtSync)}
 	case
 		// tokens that may start an expression
 		token.IDENT, token.INT, token.FLOAT, token.IMAG, token.CHAR, token.STRING, token.FUNC, token.LPAREN, // operands
@@ -1916,7 +1956,9 @@ func (p *parser) parseStmt() (s ast.Stmt) {
 		// no statement found
 		pos := p.pos
 		p.errorExpected(pos, "statement")
-		p.next() // make progress
+		for !isStmtSync(p.tok) {
+			p.next() // make progress
+		}
 		s = &ast.BadStmt{From: pos, To: p.pos}
 	}
 
@@ -2107,8 +2149,13 @@ func (p *parser) parseReceiver(scope *ast.Scope) *ast.FieldList {
 	recv := par.List[0]
 	base := deref(recv.Type)
 	if _, isIdent := base.(*ast.Ident); !isIdent {
-		p.errorExpected(base.Pos(), "(unqualified) identifier")
-		par.List = []*ast.Field{{Type: &ast.BadExpr{From: recv.Pos(), To: recv.End()}}}
+		if _, isBad := base.(*ast.BadExpr); !isBad {
+			// only report error if it's a new one
+			p.errorExpected(base.Pos(), "(unqualified) identifier")
+		}
+		par.List = []*ast.Field{
+			{Type: &ast.BadExpr{From: recv.Pos(), To: recv.End()}},
+		}
 	}
 
 	return par
@@ -2164,7 +2211,7 @@ func (p *parser) parseFuncDecl() *ast.FuncDecl {
 	return decl
 }
 
-func (p *parser) parseDecl() ast.Decl {
+func (p *parser) parseDecl(isSync func(token.Token) bool) ast.Decl {
 	if p.trace {
 		defer un(trace(p, "Declaration"))
 	}
@@ -2186,9 +2233,10 @@ func (p *parser) parseDecl() ast.Decl {
 	default:
 		pos := p.pos
 		p.errorExpected(pos, "declaration")
-		p.next() // make progress
-		decl := &ast.BadDecl{From: pos, To: p.pos}
-		return decl
+		for !isSync(p.tok) {
+			p.next() // make progress
+		}
+		return &ast.BadDecl{From: pos, To: p.pos}
 	}
 
 	return p.parseGenDecl(p.tok, f)
@@ -2227,7 +2275,7 @@ func (p *parser) parseFile() *ast.File {
 		if p.mode&ImportsOnly == 0 {
 			// rest of package body
 			for p.tok != token.EOF {
-				decls = append(decls, p.parseDecl())
+				decls = append(decls, p.parseDecl(isDeclSync))
 			}
 		}
 	}

src/pkg/go/parser/testdata/commas.src

@@ -0,0 +1,19 @@
+// Copyright 2012 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.
+
+// Test case for error messages/parser synchronization
+// after missing commas.
+
+package p
+
+var _ = []int{
+	0 /* ERROR "missing ','" */
+}
+
+var _ = []int{
+	0,
+	1,
+	2,
+	3 /* ERROR "missing ','" */
+}

src/pkg/go/parser/testdata/issue3106.src

@@ -0,0 +1,46 @@
+// Copyright 2012 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.
+
+// Test case for issue 3106: Better synchronization of
+// parser after certain syntax errors.
+
+package main
+
+func f() {
+	var m Mutex
+	c := MakeCond(&m)
+	percent := 0
+	const step = 10
+	for i := 0; i < 5; i++ {
+		go func() {
+			for {
+				// Emulates some useful work.
+				time.Sleep(1e8)
+				m.Lock()
+				defer
+				if /* ERROR "expected operand, found 'if'" */ percent == 100 {
+					m.Unlock()
+					break
+				}
+				percent++
+				if percent % step == 0 {
+					//c.Signal()
+				}
+				m.Unlock()
+			}
+		}()
+	}
+	for {
+		m.Lock()
+		if percent == 0 || percent % step != 0 {
+			c.Wait()
+		}
+		fmt.Print(",")
+		if percent == 100 {
+			m.Unlock()
+			break
+		}
+		m.Unlock()
+	}
+}