dyn RNG

[ounsworth]

Things that require randomness can now be handed an instance of &mut dyn RNG, including allowing users of the library to impl their own RNG and use it, for example, as the entropy source for an ML-DSA keygen.

[ounsworth]

@officialfrancismendoza Does b95fed9 address your first comment here?

[ounsworth]

A note on this change to traits.rs:

OLD

    fn add_seed_keymaterial(
        &mut self,
        additional_seed: impl KeyMaterialTrait,

NEW

    fn add_seed_keymaterial(
        &mut self,
        additional_seed: &dyn KeyMaterialTrait,

It was nice that this was a move and consumed the additional_seed in order to make it difficult for the user to accidentally use it twice, unfortunately, having a method that takes a moved KeyMaterialTrait breaks dyn-compatability of the RNG trait, so this has to be a borrow now. See this comment for more detail.

[officialfrancismendoza]

@officialfrancismendoza Does b95fed9 address your first comment here?

@ounsworth yes, this addresses my first comment functionally. However (nitpick), I don’t see a regression test specifically where encaps_from_rng() rejects an insufficient-strength RNG; the tests added seem focused on keygen. Did we want to add that just in case? If not, everything else looks good.

[ounsworth]

@officialfrancismendoza Oh that's not a nitpick, that's a great catch! Thanks for double-checking my work carefully!

[officialfrancismendoza]

@ounsworth looks good. Just for the record: I made sure to push upstream to source (David's bare metal server), as changes on this public mirror are overwritten on a regular interval. Since that's now reflected in 0.1.2alpha, approving this feature branch.

[officialfrancismendoza]

Approved and squash-merged to 0.1.2alpha on source (to reflect on the public mirror here). LGTM!