[Bug tree-optimization/119085] [12/13/14/15 regression] tree-sra generates wrong code for unions containing structs with holes (Emacs crashes with -O3) since r10-6321-g636e80eea24b78

[luangruo at yahoo dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119085

--- Comment #4 from Po Lu <luangruo at yahoo dot com> ---
"rguenth at gcc dot gnu.org" <gcc-bugzi...@gcc.gnu.org> writes:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119085
>
> Richard Biener <rguenth at gcc dot gnu.org> changed:
>
>            What    |Removed                     |Added
> ----------------------------------------------------------------------------
>      Ever confirmed|0                           |1
>                  CC|                            |rguenth at gcc dot gnu.org
>            Priority|P3                          |P2
>              Status|UNCONFIRMED                 |NEW
>    Last reconfirmed|                            |2025-03-02
>
> --- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
> Total scalarization of a union looks dubious unless it uses ref-all and of
> course copies the whole storage.

The union type is never considered at all because the access to the
union is of the same size as an access to a member struct, which struct
was defined first and therefore sorts before the union, with the result
that the former access is excluded by this loop in
sort_and_splice_var_accesses:


      j = i + 1;
      while (j < access_count)
        {
          struct access *ac2 = (*access_vec)[j];
          if (ac2->offset != access->offset || ac2->size != access->size)
            break;
[...]
          ac2->group_representative = access;
          j++;
        }

      i = j;