[Bug c++/42357] another ice in create_tmp_var

[rguenth at gcc dot gnu dot org]

------- Comment #2 from rguenth at gcc dot gnu dot org  2009-12-13 15:19 -------
The type is addressable and we're creating the temporary on behalf of
early SRA through force_gimple_operand_gsi.

VIEW_CONVERT_EXPR<struct PixelARGB>(SR.9075)

it's not necessary to do this gimplification as the stmt is a struct copy
before:

2449          if (force_gimple_rhs)
2450            rhs = force_gimple_operand_gsi (gsi, rhs, true, NULL_TREE,
2451                                            true, GSI_SAME_STMT);
2452          if (gimple_assign_rhs1 (*stmt) != rhs)
2453            {
2454              gimple_assign_set_rhs_from_tree (gsi, rhs);
(gdb) call debug_gimple_stmt (*stmt)
# .MEM_199 = VDEF <.MEM_91>
pixel.argb = D.93455.argb;

in fact, why the following looks wrong:

          if (!useless_type_conversion_p (TREE_TYPE (lhs), TREE_TYPE (rhs)))
            {
              rhs = fold_build1_loc (loc, VIEW_CONVERT_EXPR, TREE_TYPE (lhs),
rhs);
              if (!is_gimple_reg (lhs))
                force_gimple_rhs = true;

it should check for is_gimple_reg_type (TREE_TYPE (lhs)) instead.

I'm reducing the testcase and check a patch to that effect.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |rguenth at gcc dot gnu dot
                   |dot org                     |org
             Status|UNCONFIRMED                 |ASSIGNED
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2009-12-13 15:19:50
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42357