Update hashes to 0.16.0

[rustaceanrob]

The 0.15.0 was a massive release for hashes in terms of API clean up and also adding new HKDF support. ElligatorSwift in combination with HKDF are the primitives required for encrypted message passing on bitcoin, so exporting these in tandum would save an additional hashes dependency.

I skip to 0.16.0 because hashes allows for hex-conservative as an optional dependency, and no listed changes elsewhere.

Reducing the scope of #783 and will take a stab at rand later.

[apoelstra]

Thanks! Yeah, let's do this. I am tempted to cut a release of rust-secp256k1 anyway since @stevenroose also asked for this.

Was hoping to get #806 in first but it's blocked on lack of review, so let's just do it.

[apoelstra]

cargo test --no-default-features --features 'alloc hashes'

fails claiming that Sha256 does not implement Display.

I'm not sure why the Display impl on Sha256 should depend on anything. I spent a couple minutes trying to trace through the macro soup that produces these impls but gave up.

[rustaceanrob]

Well, I came up with a solution lol. Using the Debug string works. This macro is implementing Debug, so this seems acceptable.

[apoelstra]

Hah, thanks! I'd prefer to keep the PR here as-is as a canary, because I suspect that actually we broke hashes for nostd users, and we should fix that and cut a point release.

[rustaceanrob]

Exploring hashes, it looks like the hash_trait_impls macro (internal_macros.rs) calls the impl_hex_string_traits macro (macros.rs) conditional on the hex feature. The 0.16.0 version makes hex optional, which is perhaps explaining the failure. Displaying a secret could can be gated on a hex feature or come with std, unless you have a proposed patch for hashes.

[tcharding]

I did #833 while reviewing this.

[tcharding]

That 16 width was so its a fingerprint not the whole hash. But I cannot find a test for it. Can you add this and make it pass?

#[cfg(test)]
mod test {
    use super::*;

    #[test]
    #[cfg(feature = "hashes")]
    fn debug_secret_hashes_feature_enabled() {
        let key = "0000000000000000000000000000000000000000000000000000000000000001".parse::<SecretKey>().unwrap();
        
        // Normal debug hides value (`Display` is not implemented for `SecretKey`).
        let want = "SecretKey(#2518682f7819fb2d)";
        let got = format!("{:?}", key);
        assert_eq!(got, want);
    }

    #[test]
    #[cfg(not(feature = "hashes"))]
    fn debug_secret_no_hashes() {
        let key = "0000000000000000000000000000000000000000000000000000000000000001".parse::<SecretKey>().unwrap();
        
        // With "hashes" feature we just output a message.
        let want = "<secret key; enable `hashes` feature of `secp256k1` to display fingerprint>";
        let got = format!("{:?}", key);
        assert_eq!(got, want);
    }
}

[tcharding]

Maybe the doc test was supposed to test it but it just includes a comment about it.

[rustaceanrob]

Rebasing on #833 works. I will wait for that to go in and then add the tests.