spec: go not as smart about log.Panicf() as it is about panic()

[ellisonch]

Go is quite happy with the following program:

package main

import "fmt"
// import "log"

func foo(x int) int {
    if x == 0 {
        return 0
    }
    // log.Panicf("Not yet handling case %d", x)
    panic(fmt.Sprintf("Not yet handling case %d", x))
}

func main() {
    foo(0)
}

It understands that panic() is a "NoReturn" function, and so it eliminates foo()'s responsibility of returning an int in that case. It compiles and runs. However, if you swap the panic() with the log.Panicf() line (and of course swap the imports), go now complains:

.\panic.go:12: missing return at end of function

It seems that something needs to inform the compiler that log.Panicf() (and other similar functions) are also NoReturn.

$ go version
go version go1.4.1 windows/amd64

[bradfitz]

This is working as defined by the spec: http://golang.org/ref/spec#Terminating_statements

[ellisonch]

Where should I post issues about the spec? It seems like having the concept of NoReturn would be a nice extension of the terminating statement idea.

[bradfitz]

Generally language change requests are first discussed at https://groups.google.com/forum/#!forum/golang-nuts