Release 2.1.0

[stevenroose]

I added in one change that I had pending, I can split that up if people don't like it.

[tcharding]

Looks good to me, ACK 7416986

[stevenroose]

Gonna await another ack. @jeandudey maybe?

[jeandudey]

@stevenroose LGTM, ACK 7416986

[benma]

Please also bump the bitcoin-hashes version to align with the one published by rust-bitcoin, which is currently bitcoin-hashes=0.12.0. This way if one uses both crates, the hashes dependency is not duplicated.

[jeandudey]

@benma Maybe three versions could be released to keep everyone happy as I need bitcoin_hashes to be 0.13 as bitcoin_hashes takes a lot of flash space due to unrolling of the implementations.

E.g.:

2.1.0 with bitcoin_hashes@0.11
2.2.0 with bitcoin_hashes@0.12
2.3.0 with bitcoin_hashes@0.13

[benma]

@jeandudey can you elaborate? Which version of bitcoin_hashes does the unrolling and which doesn't? Do you have links to the relevant changes/discussions? It is relevant to me as we use the bitocoin-hashes crate in the BitBox02 firmware.

Could also go to bitcoin_hashes 0.13 and ask rust-bitcoin to also bump to 0.13.

[tcharding]

If no one needs 0.11 then we can use a range dependency over 12 and 13

diff --git a/Cargo.toml b/Cargo.toml
index 2970cc5..2e57a82 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -48,13 +48,13 @@ serde = { version = "1.0", default-features = false, features = [ "alloc" ], opt
 zeroize = { version = "1.5", features = ["zeroize_derive"], optional = true }
 
 # Unexported dependnecies
-bitcoin_hashes = { version = "0.11.0", default-features = false }
+bitcoin_hashes = { version = ">=0.12.0, <0.13", default-features = false }
 unicode-normalization = { version = "=0.1.22", optional = true }
 
 [dev-dependencies]
 # Enabling the "rand" feature by default to run the benches
 bip39 = { path = ".", features = ["rand"] }
-bitcoin_hashes = "0.11.0" # enable default features for test
+bitcoin_hashes = "0.13" # enable default features for test
 
 
 [package.metadata.docs.rs]
diff --git a/src/pbkdf2.rs b/src/pbkdf2.rs
index e7d3375..326ca55 100644
--- a/src/pbkdf2.rs
+++ b/src/pbkdf2.rs
@@ -112,7 +112,7 @@ pub(crate) fn pbkdf2<M>(mnemonic: M, unprefixed_salt: &[u8], c: usize, res: &mut
                        prfc.input(unprefixed_salt);
                        prfc.input(&u32_to_array_be((i + 1) as u32));
 
-                       let salt = hmac::Hmac::from_engine(prfc).into_inner();
+                       let salt = hmac::Hmac::from_engine(prfc).to_byte_array();
                        xor(chunk, &salt);
                        salt
                };
@@ -120,7 +120,7 @@ pub(crate) fn pbkdf2<M>(mnemonic: M, unprefixed_salt: &[u8], c: usize, res: &mut
                for _ in 1..c {
                        let mut prfc = prf.clone();
                        prfc.input(&salt);
-                       salt = hmac::Hmac::from_engine(prfc).into_inner();
+                       salt = hmac::Hmac::from_engine(prfc).to_byte_array();
 
                        xor(chunk, &salt);

[tcharding]

and ask rust-bitcoin to also bump to 0.13.

Upcoming rust-bitcoin v0.31.0 release depends on hashes v0.13. Latest rust-secp256k1 uses a range dependency as shown above.

[benma]

@tcharding that seems like a great option.

+bitcoin_hashes = { version = ">=0.12.0, <0.13", default-features = false }

Shouldn't that be <0.14?

[jeandudey]

@benma All of the versions unroll the rounds of sha256, upgrading the crate in a separate minor version is just a convenience for people that want to stick with one version of bitcoin_hashes.

The code is manually unrolled as seen here:

https://docs.rs/bitcoin_hashes/0.13.0/src/bitcoin_hashes/sha256.rs.html#340

We do also use bitcoin_hashes in Passport (specifically for sha256) so it takes a little bit of .text space.

That can be disabled by setting the small_hash feature but it's slower.

[tcharding]

+bitcoin_hashes = { version = ">=0.12.0, <0.13", default-features = false }

Shouldn't that be <0.14?

My bad, my intention was <=0.13, that's what we have in secp. I've not tested with a non-existent version (0.14).

[kayabaNerve]

I was about to submit a PR bumping to 0.13, yet there seems to be a bit of discussion above. Should I submit a PR with >=0.12, <=0.13 or was that not the best solution?

[tcharding]

Range dependency works well in secp, AFAIU it should be favoured anytime it works because its more flexible for users. But that is just my understanding, I don't speak for this project.

[kayabaNerve]

I opened #60.

[benma]

In the meantime I updated my project to bitcoin_hashes 0.13, so I am fine with either 0.13 or the range to include 0.12 as well. The other rust-bitcoin crates also seem to be at 0.13.

[benma]

@stevenroose wen release 😄

[stevenroose]

Waiting on #57 and #64. Then will replace this release MR with another one and wait for feedback.

[stevenroose]

Sorry for the delay. Merged the other MRs and updated the changelog. Will release when this gets merged.

[tcharding]

ACK 7416986

[tcharding]

CI fixes in #68 if you want them.

[tcharding]

ACK 125abeb