Re: [Fortran, Patch] PR 33917 - PROCEDURE(proc-interface) - ensure proc-interface is resolved

["Paul Richard Thomas" <paul dot richard dot thomas at gmail dot com>]

Tobias,

I am having an extremely pressed time at the moment and do not think
that I can reply reliably.  I just got in from Wien and am then on my
way to Barcelona in three hours.  It goes on like this for another two
weeks.  If this is still a problem then, I'll gladly take a look.

Cheers

Paul

On 10/27/07, Tobias Burnus <burnus@net-b.de> wrote:
> Hi all, hi Janus and Thomas,
>
> :ADDPATCH fortran:
>
> Currently,
>
>   ! sub be a procedure with an explicit interface
>   procedure(sub) :: p4
>   procedure(p4) :: p6
>
> fails if the symbol p6 is resolved before p4 as then
> p6->interface->attr.if_source is IFSRC_UNKNOWN.
>
> Whether it fails, depends how one walks the symbol tree. See PR33917 for
> an example which fails, unless you comment a line.
>
> The following patch simply checks whether the proc-interface symbol in
> "PROCEDURE([proc-interface])" has already a non-NULL sym->interface and,
> if yes, it uses it then instead of using "sym" itself.
>
> I think in terms of copying attributes this should be ok; however, I
> would not mind if you try to come up with a reason why this is wrong
> (wrong/missing attributes, memory leak, ...) or whether there is still a
> case for which it will (still) fail.
>
> Do you would like to see a comment above the added line or is its
> meaning clear?
>
> Built and regression tested on x86-64-linux.
> OK for the trunk?
>
> Tobias
>
> 2007-10-27  Tobias Burnus  <burnus@net-b.de>
>
>         PR fortran/33917
>         * decl.c (match_procedure_decl): If available, use directly
>         the procedure interface.
>
> 2007-10-27  Tobias Burnus  <burnus@net-b.de>
>
>         PR fortran/33917
>         * gfortran.dg/proc_decl_5.f90: New.
>
> Index: gcc/fortran/decl.c
> ===================================================================
> --- gcc/fortran/decl.c  (Revision 129680)
> +++ gcc/fortran/decl.c  (Arbeitskopie)
> @@ -3920,6 +3920,9 @@ match_procedure_decl (void)
>    /* Various interface checks.  */
>    if (proc_if)
>      {
> +      if (proc_if->interface)
> +       proc_if = proc_if->interface;
> +
>        if (proc_if->generic)
>         {
>           gfc_error ("Interface '%s' at %C may not be generic", proc_if->name);
> Index: gcc/testsuite/gfortran.dg/proc_decl_5.f90
> ===================================================================
> --- gcc/testsuite/gfortran.dg/proc_decl_5.f90   (Revision 0)
> +++ gcc/testsuite/gfortran.dg/proc_decl_5.f90   (Revision 0)
> @@ -0,0 +1,20 @@
> +! { dg-do compile }
> +! PR fortran/33917
> +!
> +! Depending in which order the symbol tree
> +! was walked in resolve, gfortran resolved
> +! p6 before p4; thus there was no explicit
> +! interface available for p4 and an error
> +! was printed.
> +!
> +program s
> +  implicit none
> +  procedure() :: q2
> +  procedure() :: q3
> +  procedure() :: q5
> +  procedure(sub) :: p4
> +  procedure(p4) :: p6
> +contains
> +  subroutine sub
> +  end subroutine
> +end program s
>
>


-- 
The knack of flying is learning how to throw yourself at the ground and miss.
       --Hitchhikers Guide to the Galaxy