cmd/compile/internal/syntax: establish principled position information

Until now, the parser set the position for each Node to the position of
the first token belonging to that node. For compatibility with the now
defunct gc parser, in many places that position information was modified
when the gcCompat flag was set (which it was, by default). Furthermore,
in some places, position information was not set at all.

This change removes the gcCompat flag and all associated code, and sets
position information for all nodes in a more principled way, as proposed
by mdempsky (see #16943 for details). Specifically, the position of a
node may not be at the very beginning of the respective production. For
instance for an Operation `a + b`, the position associated with the node
is the position of the `+`. Thus, for `a + b + c` we now get different
positions for the two additions.

This change does not pass toolstash -cmp because position information
recorded in export data and pcline tables is different. There are no
other functional changes.

Added test suite testing the position of all nodes.

Fixes #16943.

Change-Id: I3fc02bf096bc3b3d7d2fa655dfd4714a1a0eb90c
Reviewed-on: https://go-review.googlesource.com/37017
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>

src/cmd/compile/internal/syntax/nodes.go

@@ -10,9 +10,16 @@ import "cmd/internal/src"
 // Nodes
 
 type Node interface {
+	// Pos() returns the position associated with the node as follows:
+	// 1) The position of a node representing a terminal syntax production
+	//    (Name, BasicLit, etc.) is the position of the respective production
+	//    in the source.
+	// 2) The position of a node representing a non-terminal production
+	//    (IndexExpr, IfStmt, etc.) is the position of a token uniquely
+	//    associated with that production; usually the left-most one
+	//    ('[' for IndexExpr, 'if' for IfStmt, etc.)
 	Pos() src.Pos
 	aNode()
-	init(p *parser)
 }
 
 type node struct {
@@ -27,11 +34,6 @@ func (n *node) Pos() src.Pos {
 
 func (*node) aNode() {}
 
-// TODO(gri) we may be able to get rid of init here and in Node
-func (n *node) init(p *parser) {
-	n.pos = p.pos()
-}
-
 // ----------------------------------------------------------------------------
 // Files
 

src/cmd/compile/internal/syntax/nodes_test.go

+// Copyright 2017 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 syntax
+
+import (
+	"fmt"
+	"strings"
+	"testing"
+)
+
+// A test is a source code snippet of a particular node type.
+// In the snippet, a '@' indicates the position recorded by
+// the parser when creating the respective node.
+type test struct {
+	nodetyp string
+	snippet string
+}
+
+var decls = []test{
+	// The position of declarations is always the
+	// position of the first token of an individual
+	// declaration, independent of grouping.
+	{"ImportDecl", `import @"math"`},
+	{"ImportDecl", `import @mymath "math"`},
+	{"ImportDecl", `import @. "math"`},
+	{"ImportDecl", `import (@"math")`},
+	{"ImportDecl", `import (@mymath "math")`},
+	{"ImportDecl", `import (@. "math")`},
+
+	{"ConstDecl", `const @x`},
+	{"ConstDecl", `const @x = 0`},
+	{"ConstDecl", `const @x, y, z = 0, 1, 2`},
+	{"ConstDecl", `const (@x)`},
+	{"ConstDecl", `const (@x = 0)`},
+	{"ConstDecl", `const (@x, y, z = 0, 1, 2)`},
+
+	{"TypeDecl", `type @T int`},
+	{"TypeDecl", `type @T = int`},
+	{"TypeDecl", `type (@T int)`},
+	{"TypeDecl", `type (@T = int)`},
+
+	{"VarDecl", `var @x int`},
+	{"VarDecl", `var @x, y, z int`},
+	{"VarDecl", `var @x int = 0`},
+	{"VarDecl", `var @x, y, z int = 1, 2, 3`},
+	{"VarDecl", `var @x = 0`},
+	{"VarDecl", `var @x, y, z = 1, 2, 3`},
+	{"VarDecl", `var (@x int)`},
+	{"VarDecl", `var (@x, y, z int)`},
+	{"VarDecl", `var (@x int = 0)`},
+	{"VarDecl", `var (@x, y, z int = 1, 2, 3)`},
+	{"VarDecl", `var (@x = 0)`},
+	{"VarDecl", `var (@x, y, z = 1, 2, 3)`},
+
+	{"FuncDecl", `func @f() {}`},
+	{"FuncDecl", `func @(T) f() {}`},
+	{"FuncDecl", `func @(x T) f() {}`},
+}
+
+var exprs = []test{
+	// The position of an expression is the position
+	// of the left-most token that identifies the
+	// kind of expression.
+	{"Name", `@x`},
+
+	{"BasicLit", `@0`},
+	{"BasicLit", `@0x123`},
+	{"BasicLit", `@3.1415`},
+	{"BasicLit", `@.2718`},
+	{"BasicLit", `@1i`},
+	{"BasicLit", `@'a'`},
+	{"BasicLit", `@"abc"`},
+	{"BasicLit", "@`abc`"},
+
+	{"CompositeLit", `@{}`},
+	{"CompositeLit", `T@{}`},
+	{"CompositeLit", `struct{x, y int}@{}`},
+
+	{"KeyValueExpr", `"foo"@: true`},
+	{"KeyValueExpr", `"a"@: b`},
+
+	{"FuncLit", `@func (){}`},
+	{"ParenExpr", `@(x)`},
+	{"SelectorExpr", `a@.b`},
+	{"IndexExpr", `a@[i]`},
+
+	{"SliceExpr", `a@[:]`},
+	{"SliceExpr", `a@[i:]`},
+	{"SliceExpr", `a@[:j]`},
+	{"SliceExpr", `a@[i:j]`},
+	{"SliceExpr", `a@[i:j:k]`},
+
+	{"AssertExpr", `x@.(T)`},
+
+	{"Operation", `@*b`},
+	{"Operation", `@+b`},
+	{"Operation", `@-b`},
+	{"Operation", `@!b`},
+	{"Operation", `@^b`},
+	{"Operation", `@&b`},
+	{"Operation", `@<-b`},
+
+	{"Operation", `a @|| b`},
+	{"Operation", `a @&& b`},
+	{"Operation", `a @== b`},
+	{"Operation", `a @+ b`},
+	{"Operation", `a @* b`},
+
+	{"CallExpr", `f@()`},
+	{"CallExpr", `f@(x, y, z)`},
+	{"CallExpr", `obj.f@(1, 2, 3)`},
+	{"CallExpr", `func(x int) int { return x + 1 }@(y)`},
+
+	// ListExpr: tested via multi-value const/var declarations
+}
+
+var types = []test{
+	{"Operation", `@*T`},
+	{"Operation", `@*struct{}`},
+
+	{"ArrayType", `@[10]T`},
+	{"ArrayType", `@[...]T`},
+
+	{"SliceType", `@[]T`},
+	{"DotsType", `@...T`},
+	{"StructType", `@struct{}`},
+	{"InterfaceType", `@interface{}`},
+	{"FuncType", `func@()`},
+	{"MapType", `@map[T]T`},
+
+	{"ChanType", `@chan T`},
+	{"ChanType", `@chan<- T`},
+	{"ChanType", `@<-chan T`},
+}
+
+var fields = []test{
+	{"Field", `@T`},
+	{"Field", `@(T)`},
+	{"Field", `@x T`},
+	{"Field", `@x *(T)`},
+	{"Field", `@x, y, z T`},
+	{"Field", `@x, y, z (*T)`},
+}
+
+var stmts = []test{
+	{"EmptyStmt", `@;`},
+
+	{"LabeledStmt", `L@:`},
+	{"LabeledStmt", `L@: ;`},
+	{"LabeledStmt", `L@: f()`},
+
+	{"BlockStmt", `@{}`},
+
+	// The position of an ExprStmt is the position of the expression.
+	{"ExprStmt", `@<-ch`},
+	{"ExprStmt", `f@()`},
+	{"ExprStmt", `append@(s, 1, 2, 3)`},
+
+	{"SendStmt", `ch @<- x`},
+
+	{"DeclStmt", `@const x = 0`},
+	{"DeclStmt", `@const (x = 0)`},
+	{"DeclStmt", `@type T int`},
+	{"DeclStmt", `@type T = int`},
+	{"DeclStmt", `@type (T1 = int; T2 = float32)`},
+	{"DeclStmt", `@var x = 0`},
+	{"DeclStmt", `@var x, y, z int`},
+	{"DeclStmt", `@var (a, b = 1, 2)`},
+
+	{"AssignStmt", `x @= y`},
+	{"AssignStmt", `a, b, x @= 1, 2, 3`},
+	{"AssignStmt", `x @+= y`},
+	{"AssignStmt", `x @:= y`},
+	{"AssignStmt", `x, ok @:= f()`},
+	{"AssignStmt", `x@++`},
+	{"AssignStmt", `a[i]@--`},
+
+	{"BranchStmt", `@break`},
+	{"BranchStmt", `@break L`},
+	{"BranchStmt", `@continue`},
+	{"BranchStmt", `@continue L`},
+	{"BranchStmt", `@fallthrough`},
+	{"BranchStmt", `@goto L`},
+
+	{"CallStmt", `@defer f()`},
+	{"CallStmt", `@go f()`},
+
+	{"ReturnStmt", `@return`},
+	{"ReturnStmt", `@return x`},
+	{"ReturnStmt", `@return a, b, c`},
+
+	{"IfStmt", `@if cond {}`},
+	{"ForStmt", `@for {}`},
+	{"SwitchStmt", `@switch {}`},
+	{"SelectStmt", `@select {}`},
+}
+
+var ranges = []test{
+	{"RangeClause", `for @range s {}`},
+	{"RangeClause", `for _, i = @range s {}`},
+	{"RangeClause", `for x, i = @range s {}`},
+	{"RangeClause", `for _, i := @range s {}`},
+	{"RangeClause", `for x, i := @range s {}`},
+}
+
+var guards = []test{
+	{"TypeSwitchGuard", `switch x@.(type) {}`},
+	{"TypeSwitchGuard", `switch x := x@.(type) {}`},
+	{"TypeSwitchGuard", `switch a = b; x@.(type) {}`},
+	{"TypeSwitchGuard", `switch a := b; x := x@.(type) {}`},
+}
+
+var cases = []test{
+	{"CaseClause", ` switch { @case x: }`},
+	{"CaseClause", ` switch { @case x, y, z: }`},
+	{"CaseClause", ` switch { @case x == 1, y == 2: }`},
+	{"CaseClause", ` switch { @default: }`},
+}
+
+var comms = []test{
+	{"CommClause", `select { @case <-ch: }`},
+	{"CommClause", `select { @case x <- ch: }`},
+	{"CommClause", `select { @case x = <-ch: }`},
+	{"CommClause", `select { @case x := <-ch: }`},
+	{"CommClause", `select { @case x, ok = <-ch: }`},
+	{"CommClause", `select { @case x, ok := <-ch: }`},
+	{"CommClause", `select { @default: }`},
+}
+
+func TestPos(t *testing.T) {
+	// TODO(gri) Once we have a general tree walker, we can use that to find
+	// the first occurence of the respective node and we don't need to hand-
+	// extract the node for each specific kind of construct.
+
+	testPos(t, decls, "package p; ", "",
+		func(f *File) Node { return f.DeclList[0] },
+	)
+
+	// embed expressions in a composite literal so we can test key:value and naked composite literals
+	testPos(t, exprs, "package p; var _ = T{ ", " }",
+		func(f *File) Node { return f.DeclList[0].(*VarDecl).Values.(*CompositeLit).ElemList[0] },
+	)
+
+	// embed types in a function  signature so we can test ... types
+	testPos(t, types, "package p; func f(", ")",
+		func(f *File) Node { return f.DeclList[0].(*FuncDecl).Type.ParamList[0].Type },
+	)
+
+	testPos(t, fields, "package p; func f(", ")",
+		func(f *File) Node { return f.DeclList[0].(*FuncDecl).Type.ParamList[0] },
+	)
+
+	testPos(t, stmts, "package p; func _() { ", " } ",
+		func(f *File) Node { return f.DeclList[0].(*FuncDecl).Body[0] },
+	)
+
+	testPos(t, ranges, "package p; func _() { ", " } ",
+		func(f *File) Node { return f.DeclList[0].(*FuncDecl).Body[0].(*ForStmt).Init.(*RangeClause) },
+	)
+
+	testPos(t, guards, "package p; func _() { ", " } ",
+		func(f *File) Node { return f.DeclList[0].(*FuncDecl).Body[0].(*SwitchStmt).Tag.(*TypeSwitchGuard) },
+	)
+
+	testPos(t, cases, "package p; func _() { ", " } ",
+		func(f *File) Node { return f.DeclList[0].(*FuncDecl).Body[0].(*SwitchStmt).Body[0] },
+	)
+
+	testPos(t, comms, "package p; func _() { ", " } ",
+		func(f *File) Node { return f.DeclList[0].(*FuncDecl).Body[0].(*SelectStmt).Body[0] },
+	)
+}
+
+func testPos(t *testing.T, list []test, prefix, suffix string, extract func(*File) Node) {
+	for _, test := range list {
+		// complete source, compute @ position, and strip @ from source
+		src, index := stripAt(prefix + test.snippet + suffix)
+		if index < 0 {
+			t.Errorf("missing @: %s", src)
+			continue
+		}
+
+		// build syntaxt tree
+		file, err := ParseBytes(nil, []byte(src), nil, nil, 0)
+		if err != nil {
+			t.Errorf("parse error: %s: %v", src, err)
+			continue
+		}
+
+		// extract desired node
+		node := extract(file)
+		if typ := typeOf(node); typ != test.nodetyp {
+			t.Errorf("type error: %s: type = %s, want %s", src, typ, test.nodetyp)
+			continue
+		}
+
+		// verify node position with expected position as indicated by @
+		if col := int(node.Pos().Col()); col != index {
+			t.Errorf("pos error: %s: col = %d, want %d", src, col, index)
+			continue
+		}
+	}
+}
+
+func stripAt(s string) (string, int) {
+	if i := strings.Index(s, "@"); i >= 0 {
+		return s[:i] + s[i+1:], i
+	}
+	return s, -1
+}
+
+func typeOf(n Node) string {
+	const prefix = "*syntax."
+	k := fmt.Sprintf("%T", n)
+	if strings.HasPrefix(k, prefix) {
+		return k[len(prefix):]
+	}
+	return k
+}

test/escape_closure.go

@@ -55,9 +55,9 @@ func ClosureCallArgs4() {
 
 func ClosureCallArgs5() {
 	x := 0                     // ERROR "moved to heap: x"
-	sink = func(p *int) *int { // ERROR "leaking param: p to result ~r1" "func literal does not escape"
+	sink = func(p *int) *int { // ERROR "leaking param: p to result ~r1" "func literal does not escape" "\(func literal\)\(&x\) escapes to heap"
 		return p
-	}(&x) // ERROR "&x escapes to heap" "\(func literal\)\(&x\) escapes to heap"
+	}(&x) // ERROR "&x escapes to heap"
 }
 
 func ClosureCallArgs6() {
@@ -140,10 +140,10 @@ func ClosureCallArgs14() {
 func ClosureCallArgs15() {
 	x := 0                      // ERROR "moved to heap: x"
 	p := &x                     // ERROR "moved to heap: p" "&x escapes to heap"
-	sink = func(p **int) *int { // ERROR "leaking param: p to result ~r1 level=1" "func literal does not escape"
+	sink = func(p **int) *int { // ERROR "leaking param: p to result ~r1 level=1" "func literal does not escape" "\(func literal\)\(&p\) escapes to heap"
 		return *p
 		// BAD: p should not escape here
-	}(&p) // ERROR "&p escapes to heap" "\(func literal\)\(&p\) escapes to heap"
+	}(&p) // ERROR "&p escapes to heap"
 }
 
 func ClosureLeak1(s string) string { // ERROR "ClosureLeak1 s does not escape"