Solar Designer

> I've proceeded to take pieces from this PR into separate commits and am testing them in my fork of the repo. My commits so far passed tests, so pushed...

For SipHash, I see several kinds of changes here: 1. Use Linux's own implementation where available. 2. Avoid possible clash on `uint128_t`. 3. Avoid compiler warnings. 4. Make the code...

> For supporting zero-sized sections, I think we need to revise the hash function itself instead of its call sites. Actually, I see nothing to revise - it appears to...

> prints `1e924b9d737700d7`, so that's apparently the hash value for empty input with all-zero key (not initialized here). Searching for this value finds it in https://github.com/havenwood/digest-sip_hash so either it's correct...

@alexander-pick I think I'm done revising and merging what was needed from this PR. Can you please take a look and test the current revision (main branch in this repo)...

I'm not convinced we need to change anything, but if we introduce an `FMT_FALSE` like you describe (or maybe `FMT_HAS_FALSE_POSITIVES`?), then the other flag should in fact be called `FMT_HAS_COLLISIONS`....

Since we generally use OpenMP in the host-side code of OpenCL formats running significant pre-/post-processing on host, I think this gives us a current exhaustive list of those formats: ```...

IIRC, DiskCryptor is ransomware, so will probably be out of fashion or incompatible with versions in the wild or will use uncrackable passwords very soon if not already. We may...

#3762 suggests something called DiskCryptor isn't ransomware. Are these two different things of the same name, or does the ransomware actually reuse a pre-existing app that has legitimate uses? I'd...

> ```diff > #define __ARM_NEON 1 > ``` This piece is already within `#if !JOHN_NO_SIMD`, so a sufficient fix would be to do what @claudioandre-br suggested at first: > When...