[Bug testsuite/48506] ssa-ccp-17.c = g.i fails

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48506

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2011.04.08 08:44:22
     Ever Confirmed|0                           |1

--- Comment #4 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-04-08 08:44:22 UTC ---
Hmm, it _should_ be able to fold g.i to 0.  It doesn't because
targetm.binds_local_p () returns false on g because its DECL_COMMON.
Maybe not so on darwin?

6779      /* Uninitialized COMMON variable may be unified with symbols
6780         resolved from other modules.  */
6781      else if (DECL_COMMON (exp)
6782               && !resolved_locally
6783               && (DECL_INITIAL (exp) == NULL
6784                   || DECL_INITIAL (exp) == error_mark_node))
6785        local_p = false;

thus -fno-common makes the testcase fail for me.

I'd say change it to have an explicit initializer,

const struct Foo g = { 0 };

and scan for "return 0;" instead.