[Bug middle-end/20644] bogus uninitialized warning on unused variable

[manu at gcc dot gnu dot org]

------- Comment #7 from manu at gcc dot gnu dot org  2007-08-20 14:18 -------
Even simpler testcase:

int foo ()
{
    int i = 0;
    int j;

    if (1 == i)
        return j;

    return 0;
}

This will only be reliably fixed by building a better SSA representation.
Moving the passes around will just solve it by chance (because CCP will assume
that j undefined value is actually 0, and thus remove j). Also, it will silence
many warnings (for the same reason, CCP happily initializing uninitialized
variables)

So instead of:

foo ()
{
  int j;
  int i;
  int D.1280;

<bb 0>:
  [pr20644.c : 3] i_2 = 0;
  [pr20644.c : 6] if ([pr20644.c : 6] i_2 == 1) goto <L0>; else goto <L1>;

<L0>:;
  [pr20644.c : 7] D.1280_6 = j_5;
  [pr20644.c : 7] goto <bb 3> (<L2>);

<L1>:;
  [pr20644.c : 9] D.1280_4 = 0;

  # D.1280_1 = PHI <D.1280_6(1), D.1280_4(2)>;
<L2>:;
  return D.1280_1;

}


We could generate:

foo ()
{
  int j;
  int i;

<bb 0>:
  [pr20644.c : 3] i_2 = 0;
  [pr20644.c : 6] if ([pr20644.c : 6] i_2 == 1) goto <L0>; else goto <L1>;

<L0>:;
  [pr20644.c : 7] goto <bb 3> (<L2>);

<L1>:;
  [pr20644.c : 9] j_6 = 0;

  # j_7 = PHI <j_5(D), j_6(2)>;
<L2>:;
  return j_7;

}


-- 

manu at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |manu at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20644