Hash instances derived from Hash[Long] (e.g. Hash[List[Long]]) do not pass laws

[joroKr21]

The failing law is sameAsUniversalHash:

List(-1L).hashCode // -1698283412
Hash[List[Long]].hash(List(-1L)) // 1918838495
-1L.hashCode // 0
-1L.## // -1

Took me a while to realize this while working on Hash derivation over at kittens.

To be honest I'm not sure what's the point of that law.
Why would Hash exist if it's bound to be the same as hashCode?

[ctongfei]

I'm the author who wrote that law.

The law sameAsUniversalHash is supposed to guarantee for common types and case classes, the derived Hash instances behave in the same way as .hashCode (Not for user-defined Hash instance s).

The standard library's .hashCode implementation for Lists can be found at https://github.com/scala/scala/blob/2.12.x/src/library/scala/util/hashing/MurmurHash3.scala#L110. Note that they use .## instead of .hashCode. In the original PR for Hash (#1712), the community decided to use .hashCode instead of .## for performance reasons. So I would say that this behavior is expected.

So to move forward I think that sameAsUniversalHash should only be used internally to test Hash -- it should not be a requirement for any future Hash instances.

[joroKr21]

Makes sense to me 👍

[ctongfei]

@joroKr21 So for your work in kittens, you should just test the hashCompatibility property.

[joroKr21]

This should be addressed by #4319