KangarooTwelve

[MarekKnapek]

Checklist

• documentation is added or updated
• tests are added or updated

[sjaeckel]

I guess you're missing something there.

hash_state is a union, i.e. it would be valid to do the following:

struct sha3_state st;
sha3_shake_init((hash_state*)&st, 128);

https://godbolt.org/z/Wq3oaETcE

Please revert this change, I'd prefer to keep the codebase of all the algorithms similar.

[MarekKnapek]

Yes, that works, but. I need to keep the two SHAKE states inside a KT state between invocations of the init/process/done functions. And because of the union thing, the KT state would be unnecessary big. Keeping state for other hashes that we don't care about inside KT logic. Do you prefer me reverting this or changing every other algorithm to use its own state instead?

// Edit: Oh, I misunderstood. Now I get it. I will do the change you propose.

[sjaeckel]

Thanks a lot!

Very good PR :)

Thanks for updating the docs ❤️

What do you think of my suggestions?

[sjaeckel]

I overlooked that at the first review.

Thanks for reverting the first commit, the changes look a lot better now.

[sjaeckel]

Can one also append customization bytes in multiple calls?

I'm just asking.

And is the result of kt_customization("foobar"); the same as of kt_customization("foo"); kt_customization("bar");?

[sjaeckel]

OK, so the RFC specifies that we MAY propose an incremental interface, for either M or C, so we should also document that somewhere what we support.

I just had a look at the RFC for the first time ... the D isn't fixed to 0x1F, maybe it'd make sense to add support for variable D in a future PR. But now we focus on KT.

[MarekKnapek]

Yes, the customization could be appended multiple times. And yes, about the foobar==foo+bar. You can append unlimited amount of customization bytes, up to unsigned long count, if you want to append more, we should change the variable type holding the number of customization bytes added. To 128 bit number or even larger one. The project https://github.com/kerukuro/digestpp is quite limited in this regard, it holds the entire customization string in memory at once. My implementation holds only the SHAKE state. They also hold a 8 kB buffer, I hold a SHAKE state instead.

[sjaeckel]

Sorry for those nitpicks and thanks for your patience.

[MarekKnapek]

Sorry for those nitpicks and thanks for your patience.

No need to apologize, it is your project, that means your rules and standards. I'm not familiar with your rules, by telling me what I did wrong I'm learning them.

[sjaeckel]

That was quick! Thanks for the PR and the reactive way of handling the reviews :)