1
0
mirror of https://github.com/golang/go synced 2024-11-19 22:04:44 -07:00

cmd/vet: skip unreachable "if" and "case" code in shift check.

Such dead code is legitimate when dealing with arch-specific
types (int, uint, uintptr).

The CL removes the majority of 'too small for shift' false positives
from such a code.

Change-Id: I62c5635a1d3774ab2d71d3d7056f0589f214cbe5
Reviewed-on: https://go-review.googlesource.com/38065
Run-TryBot: Rob Pike <r@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
This commit is contained in:
Aliaksandr Valialkin 2017-03-13 19:17:41 +02:00 committed by Rob Pike
parent 6b0bd51c1c
commit 78510bd17c
6 changed files with 186 additions and 29 deletions

View File

@ -1,16 +0,0 @@
// 32bit-specific vet whitelist. See readme.txt for details.
// TODO: fix these warnings after the CL 37950 .
math/big/float.go: x[i] (32 bits) too small for shift of 32
math/big/nat.go: Word(rand.Uint32()) (32 bits) too small for shift of 32
runtime/malloc.go: uintptr(i) (32 bits) too small for shift of 40
runtime/malloc.go: uintptr(i) (32 bits) too small for shift of 40
runtime/malloc.go: uintptr(i) (32 bits) too small for shift of 40
sync/atomic/atomic_test.go: uintptr(seed + i) (32 bits) too small for shift of 32
sync/atomic/atomic_test.go: uintptr(seed+i) << 32 (32 bits) too small for shift of 32
sync/atomic/atomic_test.go: uintptr(seed + i) (32 bits) too small for shift of 32
sync/atomic/atomic_test.go: old (32 bits) too small for shift of 32
sync/atomic/atomic_test.go: old << 32 (32 bits) too small for shift of 32
sync/atomic/atomic_test.go: old (32 bits) too small for shift of 32
sync/atomic/atomic_test.go: v (32 bits) too small for shift of 32
sync/atomic/atomic_test.go: v (32 bits) too small for shift of 32

108
src/cmd/vet/dead.go Normal file
View File

@ -0,0 +1,108 @@
// 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.
//
// Simplified dead code detector. Used for skipping certain checks
// on unreachable code (for instance, shift checks on arch-specific code).
//
package main
import (
"go/ast"
"go/constant"
)
// updateDead puts unreachable "if" and "case" nodes into f.dead.
func (f *File) updateDead(node ast.Node) {
if f.dead[node] {
// The node is already marked as dead.
return
}
switch stmt := node.(type) {
case *ast.IfStmt:
// "if" branch is dead if its condition evaluates
// to constant false.
v := f.pkg.types[stmt.Cond].Value
if v == nil {
return
}
if !constant.BoolVal(v) {
f.setDead(stmt.Body)
return
}
f.setDead(stmt.Else)
case *ast.SwitchStmt:
// Case clause with empty switch tag is dead if it evaluates
// to constant false.
if stmt.Tag == nil {
BodyLoopBool:
for _, stmt := range stmt.Body.List {
cc := stmt.(*ast.CaseClause)
if cc.List == nil {
// Skip default case.
continue
}
for _, expr := range cc.List {
v := f.pkg.types[expr].Value
if v == nil || constant.BoolVal(v) {
continue BodyLoopBool
}
}
f.setDead(cc)
}
return
}
// Case clause is dead if its constant value doesn't match
// the constant value from the switch tag.
// TODO: This handles integer comparisons only.
v := f.pkg.types[stmt.Tag].Value
if v == nil || v.Kind() != constant.Int {
return
}
tagN, ok := constant.Uint64Val(v)
if !ok {
return
}
BodyLoopInt:
for _, x := range stmt.Body.List {
cc := x.(*ast.CaseClause)
if cc.List == nil {
// Skip default case.
continue
}
for _, expr := range cc.List {
v := f.pkg.types[expr].Value
if v == nil {
continue BodyLoopInt
}
n, ok := constant.Uint64Val(v)
if !ok || tagN == n {
continue BodyLoopInt
}
}
f.setDead(cc)
}
}
}
// setDead marks the node and all the children as dead.
func (f *File) setDead(node ast.Node) {
dv := deadVisitor{
f: f,
}
ast.Walk(dv, node)
}
type deadVisitor struct {
f *File
}
func (dv deadVisitor) Visit(node ast.Node) ast.Visitor {
if node == nil {
return nil
}
dv.f.dead[node] = true
return dv
}

View File

@ -194,6 +194,9 @@ type File struct {
// Registered checkers to run.
checkers map[ast.Node][]func(*File, ast.Node)
// Unreachable nodes; can be ignored in shift check.
dead map[ast.Node]bool
}
func main() {
@ -330,7 +333,13 @@ func doPackage(directory string, names []string, basePkg *Package) *Package {
}
astFiles = append(astFiles, parsedFile)
}
files = append(files, &File{fset: fs, content: data, name: name, file: parsedFile})
files = append(files, &File{
fset: fs,
content: data,
name: name,
file: parsedFile,
dead: make(map[ast.Node]bool),
})
}
if len(astFiles) == 0 {
return nil
@ -472,6 +481,7 @@ func (f *File) walkFile(name string, file *ast.File) {
// Visit implements the ast.Visitor interface.
func (f *File) Visit(node ast.Node) ast.Visitor {
f.updateDead(node)
var key ast.Node
switch node.(type) {
case *ast.AssignStmt:

View File

@ -23,6 +23,11 @@ func init() {
}
func checkShift(f *File, node ast.Node) {
if f.dead[node] {
// Skip shift checks on unreachable nodes.
return
}
switch node := node.(type) {
case *ast.BinaryExpr:
if node.Op == token.SHL || node.Op == token.SHR {

View File

@ -107,3 +107,53 @@ func ShiftTest() {
h >>= 7 * unsafe.Alignof(h)
h >>= 8 * unsafe.Alignof(h) // ERROR "too small for shift"
}
func ShiftDeadCode() {
var i int
const iBits = 8 * unsafe.Sizeof(i)
if iBits <= 32 {
if iBits == 16 {
_ = i >> 8
} else {
_ = i >> 16
}
} else {
_ = i >> 32
}
if iBits >= 64 {
_ = i << 32
if iBits == 128 {
_ = i << 64
}
} else {
_ = i << 16
}
if iBits == 64 {
_ = i << 32
}
switch iBits {
case 128, 64:
_ = i << 32
default:
_ = i << 16
}
switch {
case iBits < 32:
_ = i << 16
case iBits > 64:
_ = i << 64
default:
_ = i << 64 // ERROR "too small for shift"
}
// Make sure other vet checks work in dead code.
if iBits == 1024 {
_ = i << 512 // OK
fmt.Printf("foo %s bar", 123) // ERROR "arg 123 for printf verb %s of wrong type: untyped int"
}
}

View File

@ -953,9 +953,12 @@ func hammerSwapUint64(addr *uint64, count int) {
}
}
const arch32 = unsafe.Sizeof(uintptr(0)) == 4
func hammerSwapUintptr64(uaddr *uint64, count int) {
// only safe when uintptr is 64-bit.
// not called on 32-bit systems.
if !arch32 {
addr := (*uintptr)(unsafe.Pointer(uaddr))
seed := int(uintptr(unsafe.Pointer(&count)))
for i := 0; i < count; i++ {
@ -965,6 +968,7 @@ func hammerSwapUintptr64(uaddr *uint64, count int) {
panic(fmt.Sprintf("SwapUintptr is not atomic: %v", old))
}
}
}
}
func hammerAddInt64(uaddr *uint64, count int) {
@ -1116,8 +1120,6 @@ func hammerStoreLoadUint64(t *testing.T, paddr unsafe.Pointer) {
func hammerStoreLoadUintptr(t *testing.T, paddr unsafe.Pointer) {
addr := (*uintptr)(paddr)
var test64 uint64 = 1 << 50
arch32 := uintptr(test64) == 0
v := LoadUintptr(addr)
new := v
if arch32 {
@ -1144,8 +1146,6 @@ func hammerStoreLoadUintptr(t *testing.T, paddr unsafe.Pointer) {
func hammerStoreLoadPointer(t *testing.T, paddr unsafe.Pointer) {
addr := (*unsafe.Pointer)(paddr)
var test64 uint64 = 1 << 50
arch32 := uintptr(test64) == 0
v := uintptr(LoadPointer(addr))
new := v
if arch32 {
@ -1398,7 +1398,7 @@ func TestUnaligned64(t *testing.T) {
switch runtime.GOARCH {
default:
if unsafe.Sizeof(int(0)) != 4 {
if !arch32 {
t.Skip("test only runs on 32-bit systems")
}
case "amd64p32":