wyhash needs some kind of initial seed filtering

[rurban]

ManDeJan/zig-wyhash#2 applies to you also.
you already updated to wyhash final 3, good, so you can use those numbers.

[JalonSolov]

@ka-weihe ^^

[ka-weihe]

Thanks @rurban, I will do this.

[ka-weihe]

I'm closing this issue, because the wyhash_seed_init function got removed from the wyhash repo. I suppose this is because this is not needed in the newest version of wyhash? Can you confirm this @rurban or @dumblob ? Please open this issue again, if this is still a problem.

Update to newest wyhash: #9945

[rurban]

Tested the latest wyhash, which is now salted with the timestamp, but the known bad seeds are still there. Still testing though...

[dumblob]

Just about now a new version of wyhash got released and the primary motivation behind releasing a new version is a solution (as measured by @rurban's smhasher) to the bad seeds and related problems.

V shall update to the newest version rather sooner than later as it relies on wyhash on quite a lot of places.

See also rurban/smhasher#243 .

[rurban]

Well, the bad seeds are now much more than before (2 up to ~20), but they cause only 1 more collision (instead of 0). Which is certainly not security relevant. Concerning is that some seeds collapse now to 0, which is bad.

The PerlinNoiseAV failures are gone. If v4 is better than v3 with bad seed protection? I'll think about it. With bad seed protection it is, without not.