https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105517
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Target| |i?86-*-*
--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
I can reproduce on x86_64 with -m32.
T (p->a1x[-1].a1);
and the both failing
T (p->a1x[DIFF_MAX].a1);
T (p->a1x[SIZE_MAX].a1);
resolve to the same value now. At FRE1 we still have
&p_74(D)->a1x[-1].a1;
and
&p_74(D)->a1x[2147483647].a1;
&p_74(D)->a1x[4294967295].a1;
the address calculations overflow in ways that make the addresses appear
the same.
