Generic hashing is prone to errors

[gurpreet-]

For some reason, when using functions such as cryptoGenericHashUpdate and cryptoGenericHashFinal the tests irregularly fail. I say irregularly because they fail 1/10 times on my local Mac environment and all the time on Linux (through SemaphoreCI).

Oddly, generic hashing works if I use key sizes KEYBYTES (which is equivalent to 32) upwards but fails for KEYBYTES_MIN.

Things I've tried:

• Changing the GenericHash.State to just byte[]. It used to be the case that Lazysodium was using a custom class to handle state, but after reading this issue, I changed all the generic hash stream functions to accept byte[] instead.
• A successful thing that I tried was changing cryptoGenericHashUpdate to go through less functions which fixed generic hashing on linux for key sizes KEYBYTES upwards.

[gurpreet-]

A solution to this problem is avoiding the use of the minimum key byte value, i.e. GenericHash.KEYBYTES_MIN.

Though, this is not an optimal solution at all as people will read the docs and then receive an error when trying to use KEYBYTES_MIN in their programs.

[gurpreet-]

GenericHash.KEYBYTES_MIN has now been commented out.

You can still use it, but it's not recommended for the issues outlined above. Thorough testing is recommended if you really want to use KEYBYTES_MIN in generic hash functions.