[Fortran,Patch] PR 45170 - Implement parsing/resolution of character deferred type parameter

[Daniel Kraft]

Hi Tobias,

Tobias Burnus wrote:
>  *PING*
> 
> I updated the patch by removing the match_type_spec part and the 
> gfortran.dg/allocate_derived_1.f90 change as those changes were applied 
> as part of Steve's patch at 
> http://gcc.gnu.org/ml/fortran/2010-10/msg00285.html
> 
> Otherwise, the patch is unchanged. Rediff attached. I have rebuild the 
> patch and will regtest it, but I do not expect any failures.
> 
> OK for the trunk?

Yes.  Two comments:

@@ -701,6 +712,7 @@ match_char_length (gfc_expr **expr)
  {
    int length;
    match m;
+  bool deferred = false;

    m = gfc_match_char ('*');
    if (m != MATCH_YES)
@@ -722,7 +734,7 @@ match_char_length (gfc_expr **expr)
    if (gfc_match_char ('(') == MATCH_NO)
      goto syntax;

-  m = char_len_param_value (expr);
+  m = char_len_param_value (expr, &deferred);
    if (m != MATCH_YES && gfc_matching_function)
      {
        gfc_undo_symbols ();

It seems that in match_char_length (unlike gfc_match_char_spec) you 
simply ignore the value of deferred...  I wonder whether that can be 
"correct" -- maybe we want a check (or even assert?) that deferred is 
false if it does not need to be handled?

+      else if (sym->ts.deferred)
+	gfc_fatal_error ("Deferred type parameter not yet supported");

Do we need both this check and the one in resolve.c?  It does not hurt, 
though...  Just curious.

Thanks for the patch (to Steve and  you)!
Daniel