CODEWITHSUNDEEP
gogoroutinerecoverdefer-keyword
Wu Yunzhou
Wu Yunzhou

Reputation: 1

difference between in recover when run goroutine in new func or not

It is about defer and recover, to catch runtime error.

version 1:

func a() {
    defer func() {
        if r := recover(); r != nil {
            fmt.Println(r)
        }
    }()
    b()
}

func b() {
    go fmt.Println([]string{}[2])
}

func main() {
    a()
    time.Sleep(1 * time.Second)
    fmt.Println("end")
}

Version 2 (Only func b() changed) :

func b() {
    go func() {
        fmt.Println([]string{}[2])
    }()
}

difference run version 1:

> go run /tmp/version1.go 
runtime error: index out of range
end

and version 2:

> go run /tmp/t.go 
panic: runtime error: index out of range

goroutine 5 [running]:
main.b.func1()
    /tmp/t.go:19 +0x109
created by main.b
    /tmp/t.go:20 +0x2b

goroutine 1 [sleep]:
time.Sleep(0x3b9aca00)
    /usr/local/go/src/runtime/time.go:59 +0xf9
main.main()
    /tmp/t.go:25 +0x29
exit status 2

why it looks different? thank you if someone can give me the detail message.

Upvotes: 0

Views: 436

Answers (1)

David Budworth
David Budworth

Reputation: 11626

In version 1, your panic happens in the main goroutine because []string{}[2] must be resolved before the goroutine can be spawned.

All arguments to a function must be resolved before it gets dispatched, and goroutines are no different.

A goroutine works basically the same as any other function call except that after the stack is setup for the call, it executes in a separate context (thread).

in your second example, you are calling a function with no args and then in that function you are panicing. Since there is no call to recover() in the goroutine, you are seeing the default panic handler.

defer/recover() can only capture a panic within the same goroutine.

If you add a defer / recover to the top of b, you will be able to capture it and do whatever (ie: println it like you did in #1)

Upvotes: 4

Related Questions