Typo in test vector comment

[riastradh]

The comment in test_vectors.json says:

For derive_key, the test input is used as the input key, and the context string is 'BLAKE3 2019-12-27 16:29:52 example context'.

BLAKE3/test_vectors/test_vectors.json

Line 2 in 80260dc

"_comment": "Each test is an input length and three outputs, one for each of the hash, keyed_hash, and derive_key modes. The input in each case is filled with a 251-byte-long repeating pattern: 0, 1, 2, ..., 249, 250, 0, 1, ... The key used with keyed_hash is the 32-byte ASCII string given in the 'key' field below. For derive_key, the test input is used as the input key, and the context string is 'BLAKE3 2019-12-27 16:29:52 example context'. (As good practice for following the security requirements of derive_key, test runners should make that context string a hardcoded constant, and we do not provided it in machine-readable form.) Outputs are encoded as hexadecimal. Each case is an extended output, and implementations should also check that the first 32 bytes match their default-length output.",

However, the test vectors were actually computed with the context string BLAKE3 2019-12-27 16:29:52 test vectors context. Citations:

BLAKE3/test_vectors/src/lib.rs

Line 32 in 2fac744

pub const TEST_CONTEXT: &str = "BLAKE3 2019-12-27 16:29:52 test vectors context";

BLAKE3/test_vectors/src/lib.rs

Lines 93 to 97 in 2fac744

	let mut derive_key_out = [0; OUTPUT_LEN];
	blake3::Hasher::new_derive_key(TEST_CONTEXT)
	.update(&input)
	.finalize_xof()
	.fill(&mut derive_key_out);

By following the source code rather than the comment, I am able to reproduce the test vectors with my independent implementation of BLAKE3.

[oconnor663]

UGGGH, I can't believe I messed this up. Thank you for catching it. Will update the comment immediately.