I updated and fixed the Neon NTT Kyber implementation today and it is now also compatible with the NIST drafts.
#495 should be solved now as well.

I have tested that everything works well on a Graviton instance, but I still want to test it on a Raspberry Pi when I am back in the office on Wednesday.

On Apple Clang I get in the aarch impls:

cc -O3 -Wall -Wextra -Wpedantic -Werror -Wmissing-prototypes -Wredundant-decls -std=c99 -I../../../common  -c -o symmetric-shake.o symmetric-shake.c
symmetric-shake.c:78:2: error: no newline at end of file [-Werror,-Wnewline-eof]
}

On Apple Clang I get in the aarch impls:

cc -O3 -Wall -Wextra -Wpedantic -Werror -Wmissing-prototypes -Wredundant-decls -std=c99 -I../../../common  -c -o symmetric-shake.o symmetric-shake.c
symmetric-shake.c:78:2: error: no newline at end of file [-Werror,-Wnewline-eof]
}

Thanks. Fixed that one. Can you test again?

Seems to work on Apple now

The PPC failures are related to a change that #500 needs to introduce

I've merged #500. This means that there are a gajillion files in conflict but should also mean that we approach green CI a bit sooner.

I talked with @cryptojedi about whether the PQCrystals standards branch and the FIPS drafts are equivalent, and there are a subtleties where they are not; Peter says they've notified NIST. To confirm: are you matching the PQCrystals standard branch or the FIPS draft?

PQClean/crypto_kem/kyber768/clean/kem.c

Line 161 in 86bc011

PQCLEAN_KYBER768_CLEAN_cmov(ss, kr, KYBER_SYMBYTES, (uint8_t) (1 - fail));

I would appreciate if someone could double check if I correctly replaced the boolean operation from https://github.com/pq-crystals/kyber/blob/standard/ref/kem.c. This went wrong a couple of times in the past.

By the way, I think it looks plausible but I am also not aware of all potential problems here. @dsprenkels could you maybe have a peek?

I talked with @cryptojedi about whether the PQCrystals standards branch and the FIPS drafts are equivalent, and there are a subtleties where they are not; Peter says they've notified NIST. To confirm: are you matching the PQCrystals standard branch or the FIPS draft?

This PR is matching the PQCrystals standard branch.

I've merged #500. This means that there are a gajillion files in conflict but should also mean that we approach green CI a bit sooner.

Thanks @thomwiggers! I have rebased on top of master just now with some pain.
I have all passing tests now on my AVX2 machine and AArch64 Linux machine.
Will turn this into a real PR now.

Ugh, it seems Astyle is not happy but only on Windows?

Weirdly enough, when I run Astyle 3.4.8 on Windows 11 (Arm64) then I do not get any astyle issues...

Thanks @thomwiggers for reviewing & merging!