`at_derivation_index` panics with "The key should not contain any wildcards at this point"

[Velnbur]

I accidentally found out that if you try to build a Descriptor with DescriptorXKey and hardened derivation path, the at_derivation_index panics with The key should not contain any wildcards at this point.

use bitcoin::{
    bip32::{ChildNumber, DerivationPath, Xpriv, Xpub},
    Network,
};
use miniscript::{
    descriptor::{DescriptorXKey, Wildcard},
    Descriptor, DescriptorPublicKey,
};

fn main() {
    let xpriv = Xpriv::new_master(Network::Regtest, &[1u8; 256]).unwrap();

    let desc =
        Descriptor::<DescriptorPublicKey>::new_pkh(DescriptorPublicKey::XPub(DescriptorXKey {
            origin: None,
            xkey: Xpub::from_priv(&secp256k1::Secp256k1::new(), &xpriv),
            derivation_path: DerivationPath::from_iter([ChildNumber::Hardened { index: 1 }]),
            wildcard: Wildcard::Unhardened,
        }))
        .unwrap();

    desc.at_derivation_index(1).unwrap(); // panics because of .expect() inside
}

It makes sense that the function "fails", because at this point at_derivation_index can't derive a descriptor, but the message seems internal and misleading regarding the true cause. I expected it to return an error

[Velnbur]

The reason is that the Descriptor::at_derivation_index calls:

rust-miniscript/src/descriptor/mod.rs

Line 674 in 04f1c58

pk.clone().at_derivation_index(self.0)

then as it sets the wildcard: Wildcard::None, it expects that DefiniteDescriptorKey::new won't fail here:

rust-miniscript/src/descriptor/key.rs

Lines 940 to 941 in 04f1c58

	Ok(DefiniteDescriptorKey::new(definite)
	.expect("The key should not contain any wildcards at this point"))

but actually it also checks:

rust-miniscript/src/descriptor/key.rs

Lines 1352 to 1354 in 04f1c58

	} else if key.has_hardened_step() {
	Err(NonDefiniteKeyError::HardenedStep)
	} else if key.is_multipath() {

I'm not sure why this .expect() exists in the first place instead of propagating error further, but fix is simple

[apoelstra]

I'm not sure why this .expect() exists in the first place instead of propagating error further, but fix is simple

Because the lines above it are supposed to get rid of any wildcards. Why isn't that happening here?

[apoelstra]

Ah, I see, you explain it here -- the "should not contain any wildcards" assertion is wrong since #830. Okay.

[Velnbur]

Oh, it returned an Option before, now I see the reason of using expect. Got it