Re: [tree-ssa] live analysis on local static functions

[law at redhat dot com]

In message <20031014164618.GB11582@atrey.karlin.mff.cuni.cz>, Jan Hubicka write
s:
 >> On Tue, 2003-10-14 at 12:10, Jan Hubicka wrote:
 >> 
 >> > ;; Function q (q)
 >> > 
 >> > q ()
 >> > {
 >> >   char * <Ub244>;
 >> > 
 >> >   {
 >> >     <Ub244> = a;
 >> >     <Ub1d0>:;;
 >> >   };
 >> >   return <Ub244>;;
 >> > }
 >> > 
 >> > so we miss the initializer.  I've tracked down the problem to fact that
 >> > variable a does not have flag used set in it's annotation.  It has, when
 >> > I disable the inlining.  It is marked in create_ssa_var_map by seeing
 >> > VUSE attached to the return statement.  With inline enabled the set also
 >> > has VUSE but it's operand is not VAR_DECL itself, but SSA_NAME
 >> > 
 >> The assignment seems to be killed by DCE, so it should not be related to
 >> the used flag.  We may be missing the use-def link between the use of
 >> 'a' and its initialization.  Could you add it as a test case and open a
 >
 >Should be there a link when the initialization happens outside the
 >function?
 >> bugzilla report?
 >> 
 >> There should be a VDEF of 'a' in that initialization.  What does
 >> -fdump-tree-all-vops show?
 >
 >No VDEFs are emit for static initailizers.
 >There is VUSE <a_1> but as I've explained earlier it points to SSA_NAME
 >so we don't set tree-used flag on the original.
The set_is_used function does know how to dive inside the SSA_NAME to get
to the VAR_DECL.  This can't be the reason for the failure.

void
set_is_used (tree t)
{
  t = get_base_symbol (t);
  var_ann (t)->used = 1;
}

Where get_base_symbol has the following:

do
  {
    ...
      switch (TREE_CODE (t))
        {
        case SSA_NAME:
          t = SSA_NAME_VAR (t);
          break;
    ...
  } while ....

Jeff