[Fortran-Experiments] Merge to r120360

[Christopher D. Rickett]

hey Tobias,

thanks for updating.  will this require me to resubmit the patches that 
are still pending?

thanks.
Chris

On Wed, 21 Feb 2007, Tobias Burnus wrote:

> Hi all,
>
> I merged the Fortran-Experiments branch to r120360 (= 2007-01-02; before
> it was r120078).
> I plan to merge further, but I want to do it in little steps to make
> sure I don't miss any conflicts.
>
> The previous branch was r121965, the new brach rev. is 122202.
>
> The merge was (in the Fortran directories) visually checked [I skimmed
> the diffs] and I did a full bootstrap and check-gfortran.
>
> The only two change I had to do to make it compile / run were the following:
>
>
> I added "case BT_VOID:" to trans-expr.c's conv_arglist_function. This is
> a new function and this "case" is in:
>
> +  if (strncmp (name, "%VAL", 4) == 0)
> +    {
> +      gfc_conv_expr (se, expr);
> +      /* %VAL converts argument to default kind.  */
> +      switch (expr->ts.type)
> +       {
> [...]
> +         /* This should have been resolved away.  */
> +         case BT_UNKNOWN: case BT_CHARACTER: case BT_DERIVED:
> +         case BT_PROCEDURE: case BT_HOLLERITH: case BT_VOID:
> +           gfc_internal_error ("Bad type in conv_arglist_function");
> +       }
>
>
> Change 2:
>
> -  if (sym->attr.is_iso_c == 1)
> +  if (sym && sym->attr.is_iso_c == 1)
>
> in resolve.c's resolve_function, otherwise I got a segmentation fault
> when running gfortran.dg/nested_modules_6.f90.
>
> Tobias
>