[PATCH v2] string: Fix OOB read on generic strncmp

[Wilco Dijkstra]

Hi Adhemerval,

What kind of readahead are you trying to test/fix here? It's not clear from the
patch... So with strcmp you can safely call strlen on either input, and similarly I
believe you can call strnlen on both inputs of strncmp without getting a crash.
That means it is possible to read ahead as long as you never read past a zero
byte or past the given size.

If you couldn't read even a single byte past the first mismatch it would be hard
to create efficient implementations, especially with wide vectors...

Cheers,
Wilco