[Bug tree-optimization/81603] Various compiler UB on very large constant offsets

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81603

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Well, yes - offset_int is basically former double_int and thus good for
addresses/offsets measured in bits.

Of course in this case why do we need the offset in bits at all? ...

That said, we're looking at an ADDR_EXPR here, so we know bitsize %
BITS_PER_UNIT == 0.  Given we reject max_size == -1 this code could have used
get_addr_base_and_unit_offset instead and avoid bit precision stuff.