psbt: confusing encoding (LE uint32) of MasterKeyFingerprint

[kallerosenbaum]

The master key fingerprint in BIP174 records are byte arrays of size 4, and in descriptors they are represented by 8 hex characters. These are the first four bytes of HASH160(pubkey).

bip32.go uses little-endian when parsing the master key fingerprint into an uint32. This confuses the user (me :-P), because one would expect Bip32Derivation.MasterKeyFingerprint to be represented as a [4]byte or at least a big-endian uint32.

I suggest using [4]byte instead, because the fingerprint is a sequence of four bytes.

If the current encoding must remain because of external dependencies, Bip32Derivation.MasterKeyFingerprint should at least be documented as being little-endian.

[kallerosenbaum]

PS. I'd be happy to make a PR for the [4]byte solution if wanted.

[onyb]

I think the confusion comes from the concatenation of the master key fingerprint with uint32 little-endian integers, in BIP174. There is no mention of the byte-ordering of the parent key fingerprint in the BIP32 spec, so assuming little-endianness is not right.

On the contrary, there is an implementation of BIP32 in the codebase which uses big-endian when parsing the fingerprint into a uint32 integer.

It seems we should be able to come up with a failing test. I don't know if the [4]byte solution is better or the big-endian uint32 one; I suggest coming up with a PR and then discuss it during the review.

[Roasbeef]

Transferred from the old btcutil repo.