I don't see how the proposed solution improves matters. Even without the second optimisation, it seems that the third could still occur and cause issues. I supp

I don't see how the proposed solution improves matters. Even without the second optimisation, it seems that the third could still occur and cause issues. I suppose the compiler must only recognise the potential for a pointer writing arbitrary data when it is first cast to an integer.