[Bug ipa/63851] [5 Regression] ipa-icf miscompiles gfortran.dg/assumed_rank_(8|9|10).f90 at -O2 and above

[fxcoudert at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63851

--- Comment #4 from Francois-Xavier Coudert <fxcoudert at gcc dot gnu.org> ---
(In reply to Martin Liška from comment #2)
> There's a pair of functions 'g' and 'h' that are proved by IPA ICF to be
> equal and thunk is created (darwin does not have alias support).

If you look at the tree dump, h's argument has restrict but not g:

h (struct array7_integer(kind=4) & restrict x)
{
  check ((struct array7_integer(kind=4) *) x);
}


g (struct array7_integer(kind=4) & x)
{
  check ((struct array7_integer(kind=4) *) x);
}


Shouldn't that disable semantic equality/equivalence?