[MobiusHorizons]

Wow, that’s crazy. Does anyone have any context on why they didn’t fix this by either disallowing NULL, or not treating the pointer as non-nullable? I’m assuming there is code that was expecting this not to error, but the combination really seems like a bug not just a sharp edge.

[int_19h]

Treating the pointer as not-nullable is precisely the point of the feature, though. By letting the compiler know that there's at least N elements there, it can do things like e.g. move that read around and even prefetch if that makes the most sense.

[jacquesm]

Indeed, at a minimum you should be able to enforce that check using a compiler flag.

[uecker]

You can add that check using -fsanitize=null (and you may want to turn the diagnostic into a run-time trap)