[Patch, Fortran] PR 63727: Checks missing for proc-pointer components: Usage as actual argument when elemental

2014-12-15 Thread Janus Weil
Hi all,

here is another small diagnostic enhancement for procedure pointer
components. Regtested on x86_64-unknown-linux-gnu. Ok for trunk?

(I'm not adding a dedicated test case, since coarray_collectives_14
already includes this case as a FIXME, which I now transformed into a
dg-error.)

Cheers,
Janus


2014-12-15  Janus Weil  ja...@gcc.gnu.org

PR fortran/63727
* resolve.c (resolve_actual_arglist): Check for elemental procedure
pointer components.


2014-12-15  Janus Weil  ja...@gcc.gnu.org

PR fortran/63727
* gfortran.dg/coarray_collectives_14.f90: Address FIXME item.
Index: gcc/fortran/resolve.c
===
--- gcc/fortran/resolve.c   (Revision 218721)
+++ gcc/fortran/resolve.c   (Arbeitskopie)
@@ -1740,6 +1740,7 @@ resolve_actual_arglist (gfc_actual_arglist *arg, p
   gfc_symbol *sym;
   gfc_symtree *parent_st;
   gfc_expr *e;
+  gfc_component *comp;
   int save_need_full_assumed_size;
   bool return_value = false;
   bool actual_arg_sav = actual_arg, first_actual_arg_sav = first_actual_arg;
@@ -1967,6 +1968,14 @@ resolve_actual_arglist (gfc_actual_arglist *arg, p
}
}
 
+  comp = gfc_get_proc_ptr_comp(e);
+  if (comp  comp-attr.elemental)
+   {
+   gfc_error (ELEMENTAL procedure pointer component %qs is not 
+  allowed as an actual argument at %L, comp-name,
+  e-where);
+   }
+
   /* Fortran 2008, C1237.  */
   if (e-expr_type == EXPR_VARIABLE  gfc_is_coindexed (e)
   gfc_has_ultimate_pointer (e))
Index: gcc/testsuite/gfortran.dg/coarray_collectives_14.f90
===
--- gcc/testsuite/gfortran.dg/coarray_collectives_14.f90(Revision 
218721)
+++ gcc/testsuite/gfortran.dg/coarray_collectives_14.f90(Arbeitskopie)
@@ -62,7 +62,7 @@ program test
   call co_reduce(caf, arg3) ! { dg-error shall have two arguments }
   call co_reduce(caf, dt%arg3) ! { dg-error shall have two arguments }
   call co_reduce(caf, elem) ! { dg-error ELEMENTAL non-INTRINSIC procedure 
'elem' is not allowed as an actual argument }
-  call co_reduce(caf, dt%elem) ! { FIXME: ELEMENTAL non-INTRINSIC procedure 
'elem' is not allowed as an actual argument }
+  call co_reduce(caf, dt%elem) ! { dg-error ELEMENTAL procedure pointer 
component 'elem' is not allowed as an actual argument }
   call co_reduce(caf, realo) ! { dg-error A argument at .1. has type 
INTEGER.4. but the function passed as OPERATOR at .2. returns REAL.4. }
   call co_reduce(caf, dt%realo) ! { dg-error A argument at .1. has type 
INTEGER.4. but the function passed as OPERATOR at .2. returns REAL.4. }
   call co_reduce(caf, int8) ! { dg-error A argument at .1. has type 
INTEGER.4. but the function passed as OPERATOR at .2. returns INTEGER.8. }


Re: [Patch, Fortran] PR 63727: Checks missing for proc-pointer components: Usage as actual argument when elemental

2014-12-15 Thread Tobias Burnus
Janus Weil wrote:
 here is another small diagnostic enhancement for procedure pointer
 components. Regtested on x86_64-unknown-linux-gnu. Ok for trunk?

Looks good to me. Thanks for the patch!

Tobias

 2014-12-15  Janus Weil  ja...@gcc.gnu.org
 
 PR fortran/63727
 * resolve.c (resolve_actual_arglist): Check for elemental procedure
 pointer components.
 
 
 2014-12-15  Janus Weil  ja...@gcc.gnu.org
 
 PR fortran/63727
 * gfortran.dg/coarray_collectives_14.f90: Address FIXME item.


Re: [Patch, Fortran] PR 63727: Checks missing for proc-pointer components: Usage as actual argument when elemental

2014-12-15 Thread Janus Weil
2014-12-15 14:49 GMT+01:00 Tobias Burnus tobias.bur...@physik.fu-berlin.de:
 Janus Weil wrote:
 here is another small diagnostic enhancement for procedure pointer
 components. Regtested on x86_64-unknown-linux-gnu. Ok for trunk?

 Looks good to me. Thanks for the patch!

Thanks for the review. Committed as r218751.

Cheers,
Janus