Possible Incompleteness

[YichiZhang0613]

In num-bigint/src/bigint.rs and num-bigint/src/biguint.rs, I think the code should check modulus != 0 and exponent >= 0/modulus != 0 as comments mentioned before use them to avoid possible panic.
num-bigint/src/bigint.rs

/// Panics if the exponent is negative or the modulus is zero.
    pub fn modpow(&self, exponent: &Self, modulus: &Self) -> Self {
        power::modpow(self, exponent, modulus)
    }

num-bigint/src/biguint.rs

/// Panics if the modulus is zero.
    pub fn modpow(&self, exponent: &Self, modulus: &Self) -> Self {
        power::modpow(self, exponent, modulus)
    }

[cuviper]

Those nested functions do check their conditions right away:

num-bigint/src/bigint/power.rs

Lines 71 to 79 in e9b204c

	pub(super) fn modpow(x: &BigInt, exponent: &BigInt, modulus: &BigInt) -> BigInt {
	assert!(
	!exponent.is_negative(),
	"negative exponentiation is not supported!"
	);
	assert!(
	!modulus.is_zero(),
	"attempt to calculate with zero modulus!"
	);

num-bigint/src/biguint/power.rs

Lines 134 to 138 in e9b204c

	pub(super) fn modpow(x: &BigUint, exponent: &BigUint, modulus: &BigUint) -> BigUint {
	assert!(
	!modulus.is_zero(),
	"attempt to calculate with zero modulus!"
	);

to avoid possible panic.

How would we avoid the panic? These functions have no other way to communicate such an error. We could add checked_modpow variants that return an Option though.

[cuviper]

I see that you have opened a few similar issues on different projects. In general, it is not considered a problem in Rust if a function panics when documented preconditions are not met -- it means that it is always considered a bug on the caller.