[Bug tree-optimization/59287] New: points-to analysis confused by union accesses

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59287

            Bug ID: 59287
           Summary: points-to analysis confused by union accesses
           Product: gcc
           Version: 4.9.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: rguenth at gcc dot gnu.org

static void
get_constraint_for_component_ref (tree t, vec<ce_s> *results,
                                  bool address_p, bool lhs_p)
{
...
  /* Handle type-punning through unions.  If we are extracting a pointer
     from a union via a possibly type-punning access that pointer
     points to anything, similar to a conversion of an integer to
     a pointer.  */
  if (!lhs_p)
    {
      tree u;
      for (u = t;
           TREE_CODE (u) == COMPONENT_REF || TREE_CODE (u) == ARRAY_REF;
           u = TREE_OPERAND (u, 0))
        if (TREE_CODE (u) == COMPONENT_REF
            && TREE_CODE (TREE_TYPE (TREE_OPERAND (u, 0))) == UNION_TYPE)
          {
            struct constraint_expr temp;

            temp.offset = 0;
            temp.var = anything_id;
            temp.type = ADDRESSOF;
            results->safe_push (temp);
            return;
          }
    }

has the adverse affect that if you have code like

  foo (TREE_CODE (t))

then it makes ANYTHING escape ... which is the worst thing that can happen.

The above shouldn't be necessary.