math/big: ProbablyPrime accepts nonpositive integers as rounds count

[ALTree]

Even though only positive integers make sense, ProbablyPrime accepts null and negative integers for the "number of Miller-Rabin rounds" parameter, and happily output true (i.e prime) when given n < 1.

package main

import "math/big"

func main() {
    n := new(big.Int)
    n.SetString("5041", 10) // 71 * 71
    println(n.ProbablyPrime(0)) // true
}

http://play.golang.org/p/vLwhlt3xsC

[griesemer]

Easy to add an extra test if that's desired and documentation is not good enough.

[ALTree]

Easy to patch but I'm not sure whether

• Leave the code, change documentation ("no Miller-Rabin test is done when n < 1")
• Panic on n < 1
• Ignore the given n and set a custom nice value (i.e. 25) on n < 1

[griesemer]

I'd document that n must be > 0 and panic with an appropriate message if it's not. That makes it clear and prevents surprises.

[minux]

Sent https://golang.org/cl/2313