[Bug tree-optimization/70804] Missed tail-call

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70804

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2016-04-27
     Ever confirmed|0                           |1

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed on x86_64.

  /* Make sure the tail invocation of this function does not refer
     to local variables.  */
  FOR_EACH_LOCAL_DECL (cfun, idx, var)
    {
      if (TREE_CODE (var) != PARM_DECL
          && auto_var_in_fn_p (var, cfun->decl)
          && (ref_maybe_used_by_stmt_p (call, var)
              || call_may_clobber_ref_p (call, var)))
        return;
    }

prevents this.

void f() ()
{
  struct B b;

  <bb 2>:
  MEM[(struct E *)&b]._vptr.E = &MEM[(void *)&_ZTV1E + 16B];
  E::destroy (&b.D.2301);
  b ={v} {CLOBBER};
  return;

here ::destroy may use/clobber the stack-local b.  Not sure if we can argue
that if after the tailcall there is a clobber of b then that is safe or
what exactly was the logic with the check.