[Bug target/79066] [7 Regression] non-PIC code generated for powerpc glibc with -fpic

[amodra at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79066

Alan Modra <amodra at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |amodra at gmail dot com

--- Comment #6 from Alan Modra <amodra at gmail dot com> ---
The problem happens in lra-contraints.c:process_address_1.  Search for
HAVE_lo_sum.  A high/lo_sum address is correctly rejected by the rs6000 backend
as improper for V4 pic here:
                  *ad.inner = gen_rtx_LO_SUM (Pmode, new_reg, addr);
                  if (! valid_address_p (ad.mode, *ad.outer, ad.as))
                    {
                      /* Try to put lo_sum into register.  */
However the code then stuffs the high/lo_sum into a reg and that of course is
seen as valid.

One possible fix that occurs to me is to disable rs6000.md:elf_high when
flag_pic.