Why is GCC 4.8.2 complaining about addition under strict overflow?

Question

Consider this code (bits.c):

#include 
#include 
#include 

static uint64_t pick_bits(unsigned          


        

Answer 1

My assumption is that gcc is inlining pick_bits and thus compiling with the knowledge that hi == lo+3 which allows it to assume that hi >= lo is always true as long as lo is low enough that lo+3 doesn't overflow.