Fix the first two test vectors of verify fail test

[siv2r]

Two vectors of the sign_verify test use different partial signatures even though their signer set, message, and nonces are the same. These are those two vectors:

Vector 1 (in valid_test_ cases):

"key_indices": [0, 1, 2],
"nonce_indices": [0, 1, 2],
"aggnonce_index": 0,
"msg_index": 0,
"signer_index": 0,
"expected": "012ABBCB52B3016AC03AD82395A1A415C48B93DEF78718E62A7A90052FE224FB"

Vector 2 (in verify_fail_test_cases):

"sig": "68537CC5234E505BD14061F8DA9E90C220A181855FD8BDB7F127BB12403B4D3B",
"key_indices": [0, 1, 2],
"nonce_indices": [0, 1, 2],
"msg_index": 0,
"signer_index": 1,
"comment": "Wrong signer"

In Vector 2, if we change the signer_index to 0 (the correct index), partial_sig_verify will still fail. Therefore, it has the incorrect partial signature. As a result, creating another vector by negating this paritalsig is also incorrect.

This PR fixes Vector 2 and the one created by negating Vector 2’s partial signature.

cc @jonasnick @real-or-random

[jonasnick]

Concept ACK, thanks @siv2r

Can confirm that the comments of vector 1 and 2 of "verify_fail_test_cases" are incorrect and that the suggested vector fixes this. Check existing and suggested vector 1 with this script:

def parse_hex(x):
    return int.from_bytes(bytes.fromhex(x), byteorder="big")

wrong_s = parse_hex("68537CC5234E505BD14061F8DA9E90C220A181855FD8BDB7F127BB12403B4D3B")
assert((n - wrong_s) == parse_hex("97AC833ADCB1AFA42EBF9E0725616F3C9A0D5B614F6FE283CEAAA37A8FFAF406"))

corr_s = parse_hex("012ABBCB52B3016AC03AD82395A1A415C48B93DEF78718E62A7A90052FE224FB")
assert((n - corr_s) == parse_hex("FED54434AD4CFE953FC527DC6A5E5BE8F6234907B7C187559557CE87A0541C46"))

I doubt it's worth incrementing the patch version of the spec but it would be nice if we could somehow reflect this in the change log, for example by adding a line to the top like

* 1.0.0 (2024-05-11):
    * Fix minor issue in two test vectors.

[siv2r]

I updated the change log as suggested. Also, the verify_error_test used the same wrong sig, so I updated those too.

[siv2r]

Addressed the change log comments

[real-or-random]

utACK 83f831b -- mod the date change (Yeah, it's better to set it to the commit date, but two days more or less don't make that big of a difference in the end :D

It took me a while to understand what's going on, but it all makes sense now. Nice catch!

[real-or-random]

utACK 508e3a6

[jonasnick]

Yeah, it's better to set it to the commit date, but two days more or less don't make that big of a difference in the end

I think the purpose of the date is to make it easier for people to look up the changes in the commit history. However, the date is wrong now anyway because github UI does not show the correct date of the commit 🤷.