https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118721
--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> --- That is not the scope of ubsan. To diagnose this generally (unless forward propagation etc. makes it clear out of bounds) you really need object known object boundaries for that. So it is more like a variant of -fsanitize=pointer-subtract, except that it wouldn't be checking the case where people do ptr1 - ptr2 in the source, but would check on pointer arithmetics the old vs. new value. Bet it would be really expensive and would need libasan support (so that it doesn't mislead people that there is a problem with pointer subtraction but with pointer arithmetics).
