1
0
mirror of https://github.com/golang/go synced 2024-11-23 20:10:08 -07:00
go/test/recover.go

588 lines
11 KiB
Go
Raw Normal View History

// run
// Copyright 2010 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 of basic recover functionality.
package main
import (
"os"
runtime, cmd/gc, cmd/ld: ignore method wrappers in recover Bug #1: Issue 5406 identified an interesting case: defer iface.M() may end up calling a wrapper that copies an indirect receiver from the iface value and then calls the real M method. That's two calls down, not just one, and so recover() == nil always in the real M method, even during a panic. [For the purposes of this entire discussion, a wrapper's implementation is a function containing an ordinary call, not the optimized tail call form that is somtimes possible. The tail call does not create a second frame, so it is already handled correctly.] Fix this bug by introducing g->panicwrap, which counts the number of bytes on current stack segment that are due to wrapper calls that should not count against the recover check. All wrapper functions must now adjust g->panicwrap up on entry and back down on exit. This adds slightly to their expense; on the x86 it is a single instruction at entry and exit; on the ARM it is three. However, the alternative is to make a call to recover depend on being able to walk the stack, which I very much want to avoid. We have enough problems walking the stack for garbage collection and profiling. Also, if performance is critical in a specific case, it is already faster to use a pointer receiver and avoid this kind of wrapper entirely. Bug #2: The old code, which did not consider the possibility of two calls, already contained a check to see if the call had split its stack and so the panic-created segment was one behind the current segment. In the wrapper case, both of the two calls might split their stacks, so the panic-created segment can be two behind the current segment. Fix this by propagating the Stktop.panic flag forward during stack splits instead of looking backward during recover. Fixes #5406. R=golang-dev, iant CC=golang-dev https://golang.org/cl/13367052
2013-09-12 12:00:16 -06:00
"reflect"
"runtime"
)
func main() {
// go.tools/ssa/interp still has:
// - some lesser bugs in recover()
// - incomplete support for reflection
interp := os.Getenv("GOSSAINTERP") != ""
test1()
test1WithClosures()
test2()
test3()
if !interp {
test4()
}
test5()
test6()
test6WithClosures()
test7()
runtime, cmd/gc, cmd/ld: ignore method wrappers in recover Bug #1: Issue 5406 identified an interesting case: defer iface.M() may end up calling a wrapper that copies an indirect receiver from the iface value and then calls the real M method. That's two calls down, not just one, and so recover() == nil always in the real M method, even during a panic. [For the purposes of this entire discussion, a wrapper's implementation is a function containing an ordinary call, not the optimized tail call form that is somtimes possible. The tail call does not create a second frame, so it is already handled correctly.] Fix this bug by introducing g->panicwrap, which counts the number of bytes on current stack segment that are due to wrapper calls that should not count against the recover check. All wrapper functions must now adjust g->panicwrap up on entry and back down on exit. This adds slightly to their expense; on the x86 it is a single instruction at entry and exit; on the ARM it is three. However, the alternative is to make a call to recover depend on being able to walk the stack, which I very much want to avoid. We have enough problems walking the stack for garbage collection and profiling. Also, if performance is critical in a specific case, it is already faster to use a pointer receiver and avoid this kind of wrapper entirely. Bug #2: The old code, which did not consider the possibility of two calls, already contained a check to see if the call had split its stack and so the panic-created segment was one behind the current segment. In the wrapper case, both of the two calls might split their stacks, so the panic-created segment can be two behind the current segment. Fix this by propagating the Stktop.panic flag forward during stack splits instead of looking backward during recover. Fixes #5406. R=golang-dev, iant CC=golang-dev https://golang.org/cl/13367052
2013-09-12 12:00:16 -06:00
test8()
test9()
if !interp {
test9reflect1()
test9reflect2()
}
runtime, cmd/gc, cmd/ld: ignore method wrappers in recover Bug #1: Issue 5406 identified an interesting case: defer iface.M() may end up calling a wrapper that copies an indirect receiver from the iface value and then calls the real M method. That's two calls down, not just one, and so recover() == nil always in the real M method, even during a panic. [For the purposes of this entire discussion, a wrapper's implementation is a function containing an ordinary call, not the optimized tail call form that is somtimes possible. The tail call does not create a second frame, so it is already handled correctly.] Fix this bug by introducing g->panicwrap, which counts the number of bytes on current stack segment that are due to wrapper calls that should not count against the recover check. All wrapper functions must now adjust g->panicwrap up on entry and back down on exit. This adds slightly to their expense; on the x86 it is a single instruction at entry and exit; on the ARM it is three. However, the alternative is to make a call to recover depend on being able to walk the stack, which I very much want to avoid. We have enough problems walking the stack for garbage collection and profiling. Also, if performance is critical in a specific case, it is already faster to use a pointer receiver and avoid this kind of wrapper entirely. Bug #2: The old code, which did not consider the possibility of two calls, already contained a check to see if the call had split its stack and so the panic-created segment was one behind the current segment. In the wrapper case, both of the two calls might split their stacks, so the panic-created segment can be two behind the current segment. Fix this by propagating the Stktop.panic flag forward during stack splits instead of looking backward during recover. Fixes #5406. R=golang-dev, iant CC=golang-dev https://golang.org/cl/13367052
2013-09-12 12:00:16 -06:00
test10()
if !interp {
test10reflect1()
test10reflect2()
}
runtime, cmd/gc, cmd/ld: ignore method wrappers in recover Bug #1: Issue 5406 identified an interesting case: defer iface.M() may end up calling a wrapper that copies an indirect receiver from the iface value and then calls the real M method. That's two calls down, not just one, and so recover() == nil always in the real M method, even during a panic. [For the purposes of this entire discussion, a wrapper's implementation is a function containing an ordinary call, not the optimized tail call form that is somtimes possible. The tail call does not create a second frame, so it is already handled correctly.] Fix this bug by introducing g->panicwrap, which counts the number of bytes on current stack segment that are due to wrapper calls that should not count against the recover check. All wrapper functions must now adjust g->panicwrap up on entry and back down on exit. This adds slightly to their expense; on the x86 it is a single instruction at entry and exit; on the ARM it is three. However, the alternative is to make a call to recover depend on being able to walk the stack, which I very much want to avoid. We have enough problems walking the stack for garbage collection and profiling. Also, if performance is critical in a specific case, it is already faster to use a pointer receiver and avoid this kind of wrapper entirely. Bug #2: The old code, which did not consider the possibility of two calls, already contained a check to see if the call had split its stack and so the panic-created segment was one behind the current segment. In the wrapper case, both of the two calls might split their stacks, so the panic-created segment can be two behind the current segment. Fix this by propagating the Stktop.panic flag forward during stack splits instead of looking backward during recover. Fixes #5406. R=golang-dev, iant CC=golang-dev https://golang.org/cl/13367052
2013-09-12 12:00:16 -06:00
test11()
if !interp {
test11reflect1()
test11reflect2()
}
test111()
runtime, cmd/gc, cmd/ld: ignore method wrappers in recover Bug #1: Issue 5406 identified an interesting case: defer iface.M() may end up calling a wrapper that copies an indirect receiver from the iface value and then calls the real M method. That's two calls down, not just one, and so recover() == nil always in the real M method, even during a panic. [For the purposes of this entire discussion, a wrapper's implementation is a function containing an ordinary call, not the optimized tail call form that is somtimes possible. The tail call does not create a second frame, so it is already handled correctly.] Fix this bug by introducing g->panicwrap, which counts the number of bytes on current stack segment that are due to wrapper calls that should not count against the recover check. All wrapper functions must now adjust g->panicwrap up on entry and back down on exit. This adds slightly to their expense; on the x86 it is a single instruction at entry and exit; on the ARM it is three. However, the alternative is to make a call to recover depend on being able to walk the stack, which I very much want to avoid. We have enough problems walking the stack for garbage collection and profiling. Also, if performance is critical in a specific case, it is already faster to use a pointer receiver and avoid this kind of wrapper entirely. Bug #2: The old code, which did not consider the possibility of two calls, already contained a check to see if the call had split its stack and so the panic-created segment was one behind the current segment. In the wrapper case, both of the two calls might split their stacks, so the panic-created segment can be two behind the current segment. Fix this by propagating the Stktop.panic flag forward during stack splits instead of looking backward during recover. Fixes #5406. R=golang-dev, iant CC=golang-dev https://golang.org/cl/13367052
2013-09-12 12:00:16 -06:00
test12()
if !interp {
test12reflect1()
test12reflect2()
}
runtime, cmd/gc, cmd/ld: ignore method wrappers in recover Bug #1: Issue 5406 identified an interesting case: defer iface.M() may end up calling a wrapper that copies an indirect receiver from the iface value and then calls the real M method. That's two calls down, not just one, and so recover() == nil always in the real M method, even during a panic. [For the purposes of this entire discussion, a wrapper's implementation is a function containing an ordinary call, not the optimized tail call form that is somtimes possible. The tail call does not create a second frame, so it is already handled correctly.] Fix this bug by introducing g->panicwrap, which counts the number of bytes on current stack segment that are due to wrapper calls that should not count against the recover check. All wrapper functions must now adjust g->panicwrap up on entry and back down on exit. This adds slightly to their expense; on the x86 it is a single instruction at entry and exit; on the ARM it is three. However, the alternative is to make a call to recover depend on being able to walk the stack, which I very much want to avoid. We have enough problems walking the stack for garbage collection and profiling. Also, if performance is critical in a specific case, it is already faster to use a pointer receiver and avoid this kind of wrapper entirely. Bug #2: The old code, which did not consider the possibility of two calls, already contained a check to see if the call had split its stack and so the panic-created segment was one behind the current segment. In the wrapper case, both of the two calls might split their stacks, so the panic-created segment can be two behind the current segment. Fix this by propagating the Stktop.panic flag forward during stack splits instead of looking backward during recover. Fixes #5406. R=golang-dev, iant CC=golang-dev https://golang.org/cl/13367052
2013-09-12 12:00:16 -06:00
test13()
if !interp {
test13reflect1()
test13reflect2()
}
runtime, cmd/gc, cmd/ld: ignore method wrappers in recover Bug #1: Issue 5406 identified an interesting case: defer iface.M() may end up calling a wrapper that copies an indirect receiver from the iface value and then calls the real M method. That's two calls down, not just one, and so recover() == nil always in the real M method, even during a panic. [For the purposes of this entire discussion, a wrapper's implementation is a function containing an ordinary call, not the optimized tail call form that is somtimes possible. The tail call does not create a second frame, so it is already handled correctly.] Fix this bug by introducing g->panicwrap, which counts the number of bytes on current stack segment that are due to wrapper calls that should not count against the recover check. All wrapper functions must now adjust g->panicwrap up on entry and back down on exit. This adds slightly to their expense; on the x86 it is a single instruction at entry and exit; on the ARM it is three. However, the alternative is to make a call to recover depend on being able to walk the stack, which I very much want to avoid. We have enough problems walking the stack for garbage collection and profiling. Also, if performance is critical in a specific case, it is already faster to use a pointer receiver and avoid this kind of wrapper entirely. Bug #2: The old code, which did not consider the possibility of two calls, already contained a check to see if the call had split its stack and so the panic-created segment was one behind the current segment. In the wrapper case, both of the two calls might split their stacks, so the panic-created segment can be two behind the current segment. Fix this by propagating the Stktop.panic flag forward during stack splits instead of looking backward during recover. Fixes #5406. R=golang-dev, iant CC=golang-dev https://golang.org/cl/13367052
2013-09-12 12:00:16 -06:00
test14()
if !interp {
test14reflect1()
test14reflect2()
test15()
test16()
}
}
func die() {
runtime.Breakpoint() // can't depend on panic
}
runtime, cmd/gc, cmd/ld: ignore method wrappers in recover Bug #1: Issue 5406 identified an interesting case: defer iface.M() may end up calling a wrapper that copies an indirect receiver from the iface value and then calls the real M method. That's two calls down, not just one, and so recover() == nil always in the real M method, even during a panic. [For the purposes of this entire discussion, a wrapper's implementation is a function containing an ordinary call, not the optimized tail call form that is somtimes possible. The tail call does not create a second frame, so it is already handled correctly.] Fix this bug by introducing g->panicwrap, which counts the number of bytes on current stack segment that are due to wrapper calls that should not count against the recover check. All wrapper functions must now adjust g->panicwrap up on entry and back down on exit. This adds slightly to their expense; on the x86 it is a single instruction at entry and exit; on the ARM it is three. However, the alternative is to make a call to recover depend on being able to walk the stack, which I very much want to avoid. We have enough problems walking the stack for garbage collection and profiling. Also, if performance is critical in a specific case, it is already faster to use a pointer receiver and avoid this kind of wrapper entirely. Bug #2: The old code, which did not consider the possibility of two calls, already contained a check to see if the call had split its stack and so the panic-created segment was one behind the current segment. In the wrapper case, both of the two calls might split their stacks, so the panic-created segment can be two behind the current segment. Fix this by propagating the Stktop.panic flag forward during stack splits instead of looking backward during recover. Fixes #5406. R=golang-dev, iant CC=golang-dev https://golang.org/cl/13367052
2013-09-12 12:00:16 -06:00
func mustRecoverBody(v1, v2, v3, x interface{}) {
v := v1
if v != nil {
println("spurious recover", v)
die()
}
v = v2
if v == nil {
println("missing recover", x.(int))
die() // panic is useless here
}
if v != x {
println("wrong value", v, x)
die()
}
// the value should be gone now regardless
runtime, cmd/gc, cmd/ld: ignore method wrappers in recover Bug #1: Issue 5406 identified an interesting case: defer iface.M() may end up calling a wrapper that copies an indirect receiver from the iface value and then calls the real M method. That's two calls down, not just one, and so recover() == nil always in the real M method, even during a panic. [For the purposes of this entire discussion, a wrapper's implementation is a function containing an ordinary call, not the optimized tail call form that is somtimes possible. The tail call does not create a second frame, so it is already handled correctly.] Fix this bug by introducing g->panicwrap, which counts the number of bytes on current stack segment that are due to wrapper calls that should not count against the recover check. All wrapper functions must now adjust g->panicwrap up on entry and back down on exit. This adds slightly to their expense; on the x86 it is a single instruction at entry and exit; on the ARM it is three. However, the alternative is to make a call to recover depend on being able to walk the stack, which I very much want to avoid. We have enough problems walking the stack for garbage collection and profiling. Also, if performance is critical in a specific case, it is already faster to use a pointer receiver and avoid this kind of wrapper entirely. Bug #2: The old code, which did not consider the possibility of two calls, already contained a check to see if the call had split its stack and so the panic-created segment was one behind the current segment. In the wrapper case, both of the two calls might split their stacks, so the panic-created segment can be two behind the current segment. Fix this by propagating the Stktop.panic flag forward during stack splits instead of looking backward during recover. Fixes #5406. R=golang-dev, iant CC=golang-dev https://golang.org/cl/13367052
2013-09-12 12:00:16 -06:00
v = v3
if v != nil {
println("recover didn't recover")
die()
}
}
runtime, cmd/gc, cmd/ld: ignore method wrappers in recover Bug #1: Issue 5406 identified an interesting case: defer iface.M() may end up calling a wrapper that copies an indirect receiver from the iface value and then calls the real M method. That's two calls down, not just one, and so recover() == nil always in the real M method, even during a panic. [For the purposes of this entire discussion, a wrapper's implementation is a function containing an ordinary call, not the optimized tail call form that is somtimes possible. The tail call does not create a second frame, so it is already handled correctly.] Fix this bug by introducing g->panicwrap, which counts the number of bytes on current stack segment that are due to wrapper calls that should not count against the recover check. All wrapper functions must now adjust g->panicwrap up on entry and back down on exit. This adds slightly to their expense; on the x86 it is a single instruction at entry and exit; on the ARM it is three. However, the alternative is to make a call to recover depend on being able to walk the stack, which I very much want to avoid. We have enough problems walking the stack for garbage collection and profiling. Also, if performance is critical in a specific case, it is already faster to use a pointer receiver and avoid this kind of wrapper entirely. Bug #2: The old code, which did not consider the possibility of two calls, already contained a check to see if the call had split its stack and so the panic-created segment was one behind the current segment. In the wrapper case, both of the two calls might split their stacks, so the panic-created segment can be two behind the current segment. Fix this by propagating the Stktop.panic flag forward during stack splits instead of looking backward during recover. Fixes #5406. R=golang-dev, iant CC=golang-dev https://golang.org/cl/13367052
2013-09-12 12:00:16 -06:00
func doubleRecover() interface{} {
return recover()
}
func mustRecover(x interface{}) {
mustRecoverBody(doubleRecover(), recover(), recover(), x)
}
func mustNotRecover() {
v := recover()
if v != nil {
println("spurious recover", v)
die()
}
}
func withoutRecover() {
mustNotRecover() // because it's a sub-call
}
func withoutRecoverRecursive(n int) {
if n == 0 {
withoutRecoverRecursive(1)
} else {
v := recover()
if v != nil {
println("spurious recover (recursive)", v)
die()
}
}
}
func test1() {
defer mustNotRecover() // because mustRecover will squelch it
defer mustRecover(1) // because of panic below
defer withoutRecover() // should be no-op, leaving for mustRecover to find
defer withoutRecoverRecursive(0) // ditto
panic(1)
}
// Repeat test1 with closures instead of standard function.
// Interesting because recover bases its decision
// on the frame pointer of its caller, and a closure's
// frame pointer is in the middle of its actual arguments
// (after the hidden ones for the closed-over variables).
func test1WithClosures() {
defer func() {
v := recover()
if v != nil {
println("spurious recover in closure")
die()
}
}()
defer func(x interface{}) {
mustNotRecover()
v := recover()
if v == nil {
println("missing recover", x.(int))
die()
}
if v != x {
println("wrong value", v, x)
die()
}
}(1)
defer func() {
mustNotRecover()
}()
panic(1)
}
func test2() {
// Recover only sees the panic argument
// if it is called from a deferred call.
// It does not see the panic when called from a call within a deferred call (too late)
// nor does it see the panic when it *is* the deferred call (too early).
defer mustRecover(2)
defer recover() // should be no-op
panic(2)
}
func test3() {
defer mustNotRecover()
defer func() {
recover() // should squelch
}()
panic(3)
}
func test4() {
// Equivalent to test3 but using defer to make the call.
defer mustNotRecover()
defer func() {
defer recover() // should squelch
}()
panic(4)
}
// Check that closures can set output arguments.
// Run g(). If it panics, return x; else return deflt.
func try(g func(), deflt interface{}) (x interface{}) {
defer func() {
if v := recover(); v != nil {
x = v
}
}()
defer g()
return deflt
}
// Check that closures can set output arguments.
// Run g(). If it panics, return x; else return deflt.
func try1(g func(), deflt interface{}) (x interface{}) {
defer func() {
if v := recover(); v != nil {
x = v
}
}()
defer g()
x = deflt
return
}
func test5() {
v := try(func() { panic(5) }, 55).(int)
if v != 5 {
println("wrong value", v, 5)
die()
}
s := try(func() {}, "hi").(string)
if s != "hi" {
println("wrong value", s, "hi")
die()
}
v = try1(func() { panic(5) }, 55).(int)
if v != 5 {
println("try1 wrong value", v, 5)
die()
}
s = try1(func() {}, "hi").(string)
if s != "hi" {
println("try1 wrong value", s, "hi")
die()
}
}
// When a deferred big call starts, it must first
// create yet another stack segment to hold the
// giant frame for x. Make sure that doesn't
// confuse recover.
func big(mustRecover bool) {
var x [100000]int
x[0] = 1
x[99999] = 1
_ = x
v := recover()
if mustRecover {
if v == nil {
println("missing big recover")
die()
}
} else {
if v != nil {
println("spurious big recover")
die()
}
}
}
func test6() {
defer big(false)
defer big(true)
panic(6)
}
func test6WithClosures() {
defer func() {
var x [100000]int
x[0] = 1
x[99999] = 1
_ = x
if recover() != nil {
println("spurious big closure recover")
die()
}
}()
defer func() {
var x [100000]int
x[0] = 1
x[99999] = 1
_ = x
if recover() == nil {
println("missing big closure recover")
die()
}
}()
panic("6WithClosures")
}
func test7() {
ok := false
func() {
// should panic, then call mustRecover 7, which stops the panic.
// then should keep processing ordinary defers earlier than that one
// before returning.
// this test checks that the defer func on the next line actually runs.
defer func() { ok = true }()
defer mustRecover(7)
panic(7)
}()
if !ok {
println("did not run ok func")
die()
}
}
func varargs(s *int, a ...int) {
*s = 0
for _, v := range a {
*s += v
}
if recover() != nil {
*s += 100
}
}
func test8a() (r int) {
defer varargs(&r, 1, 2, 3)
panic(0)
}
func test8b() (r int) {
defer varargs(&r, 4, 5, 6)
return
}
func test8() {
if test8a() != 106 || test8b() != 15 {
println("wrong value")
die()
}
}
runtime, cmd/gc, cmd/ld: ignore method wrappers in recover Bug #1: Issue 5406 identified an interesting case: defer iface.M() may end up calling a wrapper that copies an indirect receiver from the iface value and then calls the real M method. That's two calls down, not just one, and so recover() == nil always in the real M method, even during a panic. [For the purposes of this entire discussion, a wrapper's implementation is a function containing an ordinary call, not the optimized tail call form that is somtimes possible. The tail call does not create a second frame, so it is already handled correctly.] Fix this bug by introducing g->panicwrap, which counts the number of bytes on current stack segment that are due to wrapper calls that should not count against the recover check. All wrapper functions must now adjust g->panicwrap up on entry and back down on exit. This adds slightly to their expense; on the x86 it is a single instruction at entry and exit; on the ARM it is three. However, the alternative is to make a call to recover depend on being able to walk the stack, which I very much want to avoid. We have enough problems walking the stack for garbage collection and profiling. Also, if performance is critical in a specific case, it is already faster to use a pointer receiver and avoid this kind of wrapper entirely. Bug #2: The old code, which did not consider the possibility of two calls, already contained a check to see if the call had split its stack and so the panic-created segment was one behind the current segment. In the wrapper case, both of the two calls might split their stacks, so the panic-created segment can be two behind the current segment. Fix this by propagating the Stktop.panic flag forward during stack splits instead of looking backward during recover. Fixes #5406. R=golang-dev, iant CC=golang-dev https://golang.org/cl/13367052
2013-09-12 12:00:16 -06:00
type I interface {
M()
}
runtime, cmd/gc, cmd/ld: ignore method wrappers in recover Bug #1: Issue 5406 identified an interesting case: defer iface.M() may end up calling a wrapper that copies an indirect receiver from the iface value and then calls the real M method. That's two calls down, not just one, and so recover() == nil always in the real M method, even during a panic. [For the purposes of this entire discussion, a wrapper's implementation is a function containing an ordinary call, not the optimized tail call form that is somtimes possible. The tail call does not create a second frame, so it is already handled correctly.] Fix this bug by introducing g->panicwrap, which counts the number of bytes on current stack segment that are due to wrapper calls that should not count against the recover check. All wrapper functions must now adjust g->panicwrap up on entry and back down on exit. This adds slightly to their expense; on the x86 it is a single instruction at entry and exit; on the ARM it is three. However, the alternative is to make a call to recover depend on being able to walk the stack, which I very much want to avoid. We have enough problems walking the stack for garbage collection and profiling. Also, if performance is critical in a specific case, it is already faster to use a pointer receiver and avoid this kind of wrapper entirely. Bug #2: The old code, which did not consider the possibility of two calls, already contained a check to see if the call had split its stack and so the panic-created segment was one behind the current segment. In the wrapper case, both of the two calls might split their stacks, so the panic-created segment can be two behind the current segment. Fix this by propagating the Stktop.panic flag forward during stack splits instead of looking backward during recover. Fixes #5406. R=golang-dev, iant CC=golang-dev https://golang.org/cl/13367052
2013-09-12 12:00:16 -06:00
// pointer receiver, so no wrapper in i.M()
type T1 struct{}
runtime, cmd/gc, cmd/ld: ignore method wrappers in recover Bug #1: Issue 5406 identified an interesting case: defer iface.M() may end up calling a wrapper that copies an indirect receiver from the iface value and then calls the real M method. That's two calls down, not just one, and so recover() == nil always in the real M method, even during a panic. [For the purposes of this entire discussion, a wrapper's implementation is a function containing an ordinary call, not the optimized tail call form that is somtimes possible. The tail call does not create a second frame, so it is already handled correctly.] Fix this bug by introducing g->panicwrap, which counts the number of bytes on current stack segment that are due to wrapper calls that should not count against the recover check. All wrapper functions must now adjust g->panicwrap up on entry and back down on exit. This adds slightly to their expense; on the x86 it is a single instruction at entry and exit; on the ARM it is three. However, the alternative is to make a call to recover depend on being able to walk the stack, which I very much want to avoid. We have enough problems walking the stack for garbage collection and profiling. Also, if performance is critical in a specific case, it is already faster to use a pointer receiver and avoid this kind of wrapper entirely. Bug #2: The old code, which did not consider the possibility of two calls, already contained a check to see if the call had split its stack and so the panic-created segment was one behind the current segment. In the wrapper case, both of the two calls might split their stacks, so the panic-created segment can be two behind the current segment. Fix this by propagating the Stktop.panic flag forward during stack splits instead of looking backward during recover. Fixes #5406. R=golang-dev, iant CC=golang-dev https://golang.org/cl/13367052
2013-09-12 12:00:16 -06:00
func (*T1) M() {
mustRecoverBody(doubleRecover(), recover(), recover(), 9)
}
func test9() {
var i I = &T1{}
defer i.M()
panic(9)
}
func test9reflect1() {
f := reflect.ValueOf(&T1{}).Method(0).Interface().(func())
defer f()
panic(9)
}
func test9reflect2() {
f := reflect.TypeOf(&T1{}).Method(0).Func.Interface().(func(*T1))
defer f(&T1{})
panic(9)
}
// word-sized value receiver, so no wrapper in i.M()
type T2 uintptr
func (T2) M() {
mustRecoverBody(doubleRecover(), recover(), recover(), 10)
}
func test10() {
var i I = T2(0)
defer i.M()
panic(10)
}
func test10reflect1() {
f := reflect.ValueOf(T2(0)).Method(0).Interface().(func())
defer f()
panic(10)
}
func test10reflect2() {
f := reflect.TypeOf(T2(0)).Method(0).Func.Interface().(func(T2))
defer f(T2(0))
panic(10)
}
// tiny receiver, so basic wrapper in i.M()
type T3 struct{}
runtime, cmd/gc, cmd/ld: ignore method wrappers in recover Bug #1: Issue 5406 identified an interesting case: defer iface.M() may end up calling a wrapper that copies an indirect receiver from the iface value and then calls the real M method. That's two calls down, not just one, and so recover() == nil always in the real M method, even during a panic. [For the purposes of this entire discussion, a wrapper's implementation is a function containing an ordinary call, not the optimized tail call form that is somtimes possible. The tail call does not create a second frame, so it is already handled correctly.] Fix this bug by introducing g->panicwrap, which counts the number of bytes on current stack segment that are due to wrapper calls that should not count against the recover check. All wrapper functions must now adjust g->panicwrap up on entry and back down on exit. This adds slightly to their expense; on the x86 it is a single instruction at entry and exit; on the ARM it is three. However, the alternative is to make a call to recover depend on being able to walk the stack, which I very much want to avoid. We have enough problems walking the stack for garbage collection and profiling. Also, if performance is critical in a specific case, it is already faster to use a pointer receiver and avoid this kind of wrapper entirely. Bug #2: The old code, which did not consider the possibility of two calls, already contained a check to see if the call had split its stack and so the panic-created segment was one behind the current segment. In the wrapper case, both of the two calls might split their stacks, so the panic-created segment can be two behind the current segment. Fix this by propagating the Stktop.panic flag forward during stack splits instead of looking backward during recover. Fixes #5406. R=golang-dev, iant CC=golang-dev https://golang.org/cl/13367052
2013-09-12 12:00:16 -06:00
func (T3) M() {
mustRecoverBody(doubleRecover(), recover(), recover(), 11)
}
func test11() {
var i I = T3{}
defer i.M()
panic(11)
}
func test11reflect1() {
f := reflect.ValueOf(T3{}).Method(0).Interface().(func())
defer f()
panic(11)
}
func test11reflect2() {
f := reflect.TypeOf(T3{}).Method(0).Func.Interface().(func(T3))
defer f(T3{})
panic(11)
}
// tiny receiver, so basic wrapper in i.M()
type T3deeper struct{}
func (T3deeper) M() {
badstate() // difference from T3
mustRecoverBody(doubleRecover(), recover(), recover(), 111)
}
func test111() {
var i I = T3deeper{}
defer i.M()
panic(111)
}
type Tiny struct{}
func (Tiny) M() {
panic(112)
}
// i.M is a wrapper, and i.M panics.
//
// This is a torture test for an old implementation of recover that
// tried to deal with wrapper functions by doing some argument
// positioning math on both entry and exit. Doing anything on exit
// is a problem because sometimes functions exit via panic instead
// of an ordinary return, so panic would have to know to do the
// same math when unwinding the stack. It gets complicated fast.
// This particular test never worked with the old scheme, because
// panic never did the right unwinding math.
//
// The new scheme adjusts Panic.argp on entry to a wrapper.
// It has no exit work, so if a wrapper is interrupted by a panic,
// there's no cleanup that panic itself must do.
// This test just works now.
func badstate() {
defer func() {
recover()
}()
var i I = Tiny{}
i.M()
}
runtime, cmd/gc, cmd/ld: ignore method wrappers in recover Bug #1: Issue 5406 identified an interesting case: defer iface.M() may end up calling a wrapper that copies an indirect receiver from the iface value and then calls the real M method. That's two calls down, not just one, and so recover() == nil always in the real M method, even during a panic. [For the purposes of this entire discussion, a wrapper's implementation is a function containing an ordinary call, not the optimized tail call form that is somtimes possible. The tail call does not create a second frame, so it is already handled correctly.] Fix this bug by introducing g->panicwrap, which counts the number of bytes on current stack segment that are due to wrapper calls that should not count against the recover check. All wrapper functions must now adjust g->panicwrap up on entry and back down on exit. This adds slightly to their expense; on the x86 it is a single instruction at entry and exit; on the ARM it is three. However, the alternative is to make a call to recover depend on being able to walk the stack, which I very much want to avoid. We have enough problems walking the stack for garbage collection and profiling. Also, if performance is critical in a specific case, it is already faster to use a pointer receiver and avoid this kind of wrapper entirely. Bug #2: The old code, which did not consider the possibility of two calls, already contained a check to see if the call had split its stack and so the panic-created segment was one behind the current segment. In the wrapper case, both of the two calls might split their stacks, so the panic-created segment can be two behind the current segment. Fix this by propagating the Stktop.panic flag forward during stack splits instead of looking backward during recover. Fixes #5406. R=golang-dev, iant CC=golang-dev https://golang.org/cl/13367052
2013-09-12 12:00:16 -06:00
// large receiver, so basic wrapper in i.M()
type T4 [2]string
func (T4) M() {
mustRecoverBody(doubleRecover(), recover(), recover(), 12)
}
func test12() {
var i I = T4{}
defer i.M()
panic(12)
}
func test12reflect1() {
f := reflect.ValueOf(T4{}).Method(0).Interface().(func())
defer f()
panic(12)
}
func test12reflect2() {
f := reflect.TypeOf(T4{}).Method(0).Func.Interface().(func(T4))
defer f(T4{})
panic(12)
}
// enormous receiver, so wrapper splits stack to call M
type T5 [8192]byte
func (T5) M() {
mustRecoverBody(doubleRecover(), recover(), recover(), 13)
}
func test13() {
var i I = T5{}
defer i.M()
panic(13)
}
func test13reflect1() {
f := reflect.ValueOf(T5{}).Method(0).Interface().(func())
defer f()
panic(13)
}
func test13reflect2() {
f := reflect.TypeOf(T5{}).Method(0).Func.Interface().(func(T5))
defer f(T5{})
panic(13)
}
// enormous receiver + enormous method frame, so wrapper splits stack to call M,
// and then M splits stack to allocate its frame.
// recover must look back two frames to find the panic.
type T6 [8192]byte
var global byte
func (T6) M() {
var x [8192]byte
x[0] = 1
x[1] = 2
for i := range x {
global += x[i]
}
mustRecoverBody(doubleRecover(), recover(), recover(), 14)
}
func test14() {
var i I = T6{}
defer i.M()
panic(14)
}
func test14reflect1() {
f := reflect.ValueOf(T6{}).Method(0).Interface().(func())
defer f()
panic(14)
}
func test14reflect2() {
f := reflect.TypeOf(T6{}).Method(0).Func.Interface().(func(T6))
defer f(T6{})
panic(14)
}
// function created by reflect.MakeFunc
func reflectFunc(args []reflect.Value) (results []reflect.Value) {
mustRecoverBody(doubleRecover(), recover(), recover(), 15)
return nil
}
func test15() {
f := reflect.MakeFunc(reflect.TypeOf((func())(nil)), reflectFunc).Interface().(func())
defer f()
panic(15)
}
func reflectFunc2(args []reflect.Value) (results []reflect.Value) {
// This will call reflectFunc3
args[0].Interface().(func())()
return nil
}
func reflectFunc3(args []reflect.Value) (results []reflect.Value) {
if v := recover(); v != nil {
println("spurious recover", v)
die()
}
return nil
}
func test16() {
defer mustRecover(16)
f2 := reflect.MakeFunc(reflect.TypeOf((func(func()))(nil)), reflectFunc2).Interface().(func(func()))
f3 := reflect.MakeFunc(reflect.TypeOf((func())(nil)), reflectFunc3).Interface().(func())
defer f2(f3)
panic(16)
}