[Bug c++/82101] Uninitialized variable detection failed with variables in for loop

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82101

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |diagnostic,
                   |                            |missed-optimization
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2017-09-05
             Blocks|                            |24639
     Ever confirmed|0                           |1

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Dup of some existing bug for sure.  t looks like initialized to GCC on the
backedge:

  <bb 3> [0.00%]:
  # i_1 = PHI <i_5(2), i_8(4)>
  # t_2 = PHI <t_6(D)(2), t_7(4)>
  if (i_1 > 9)
    goto <bb 5>; [0.00%]
  else
    goto <bb 4>; [0.00%]

  <bb 4> [0.00%]:
  t_7 = t_2;
  i_8 = i_1 + 1;
  goto <bb 3>; [0.00%]

but that gets later optimistically propagated away (the variable is dead
anyway)
and thus the late pass warning about maybe-uninitialized doesn't get here.

Now what is special here is that the uninit value is on the loop entry edge
so the early pass _could_ warn.

It also looks like we could improve on scoping here when re-writing into SSA.
t doesn't need a PHI node as the value is not live over the backedge.


Referenced Bugs:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=24639
[Bug 24639] [meta-bug] bug to track all Wuninitialized issues