Patchwork [Fortran] PR 45586: Correct access to parent components of DT on tree level

login
register
mail settings
Submitter Tobias Burnus
Date Feb. 12, 2011, 11:20 a.m.
Message ID <4D566D1A.5060404@net-b.de>
Download mbox | patch
Permalink /patch/82900/
State New
Headers show

Comments

Tobias Burnus - Feb. 12, 2011, 11:20 a.m.
The issue was found by Michael when working on the "restrict" issue of 
DT; see http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45586#c40

Without the patch, not only his draft patch fails, but one also gets - 
as Mikael pointed out - strange dumps. For extends_1.f03 the result 
without patch is:

new_person->service.education.person.service.education.person.education.person.person.ss
= *ss;

with the patch the result is the more reasonable:

new_person->service.education.person.ss = *ss;

Build and regtested on x86-64-linux.
OK for the trunk?

Tobias

PS: Michael plans to submit his restrict patch (comment 35) after this 
patch has been committed; the draft patch has been already pre-reviewed 
by Mikael (comment 38).
Jerry DeLisle - Feb. 12, 2011, 12:08 p.m.
On 02/12/2011 03:20 AM, Tobias Burnus wrote:
> The issue was found by Michael when working on the "restrict" issue of DT; see
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45586#c40
>
> Without the patch, not only his draft patch fails, but one also gets - as Mikael
> pointed out - strange dumps. For extends_1.f03 the result without patch is:
>
> new_person->service.education.person.service.education.person.education.person.person.ss
>
> = *ss;
>
> with the patch the result is the more reasonable:
>
> new_person->service.education.person.ss = *ss;
>
> Build and regtested on x86-64-linux.
> OK for the trunk?
>
> Tobias
>
> PS: Michael plans to submit his restrict patch (comment 35) after this patch has
> been committed; the draft patch has been already pre-reviewed by Mikael (comment
> 38).

OK, thanks for patch.

Jerry
Mikael Morin - Feb. 12, 2011, 2:48 p.m.
On Saturday 12 February 2011 12:20:58 Tobias Burnus wrote:
> The issue was found by Michael when working on the "restrict" issue of
> DT; see http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45586#c40
> 
> Without the patch, not only his draft patch fails, but one also gets -
> as Mikael pointed out - strange dumps. For extends_1.f03 the result
> without patch is:
> 
> new_person->service.education.person.service.education.person.education.per
> son.person.ss = *ss;
> 
> with the patch the result is the more reasonable:
> 
> new_person->service.education.person.ss = *ss;
> 
> Build and regtested on x86-64-linux.
> OK for the trunk?
> 
> Tobias
> 
> PS: Michael plans to submit his restrict patch (comment 35) after this
> patch has been committed; the draft patch has been already pre-reviewed
> by Mikael (comment 38).

Hello, 

I'm going to commit the attached testcase.

Mikael


2011-02-12  Mikael Morin  <mikael.morin@sfr.fr>

	PR fortran/45586
	* gfortran.dg/extends_11.f03: New.
! { dg-do compile }
! { dg-options "-fdump-tree-original" }
!
! PR fortran/45586
! Test that access to inherited components are properly generated
!
! Stripped down from extends_1.f03
!
  type :: person
    integer :: ss = 1
  end type person

  type, extends(person) :: education
    integer ::  attainment = 0
  end type education

  type, extends(education) :: service
    integer :: personnel_number = 0
  end type service

  type, extends(service) :: person_record
    type (person_record), pointer :: supervisor => NULL ()
  end type person_record
  
  type(person_record) :: recruit
  

  ! Check that references by ultimate component and by parent type work
  ! All the following component access are equivalent
  recruit%ss = 2
  recruit%person%ss = 3
  recruit%education%ss = 4
  recruit%education%person%ss = 5
  recruit%service%ss = 6
  recruit%service%person%ss = 7
  recruit%service%education%ss = 8
  recruit%service%education%person%ss = 9
end

! { dg-final { scan-tree-dump-times " +recruit\\.service\\.education\\.person\\.ss =" 8 "original"} }
! { dg-final { cleanup-tree-dump "original" } }

Patch

2011-02-12  Michael Matz  <matz@suse.de>
	    Janus Weil  <janus@gcc.gnu.org>
	    Tobias Burnus  <burnus@net-b.de>

	PR fortran/45586
	* trans-expr.c (conv_parent_component_references): Avoid unintendent skipping
	of parent compounds.

diff --git a/gcc/fortran/trans-expr.c b/gcc/fortran/trans-expr.c
index f19c015..b7d7ed9 100644
--- a/gcc/fortran/trans-expr.c
+++ b/gcc/fortran/trans-expr.c
@@ -538,6 +538,11 @@  conv_parent_component_references (gfc_se * se, gfc_ref * ref)
   dt = ref->u.c.sym;
   c = ref->u.c.component;
 
+  /* Return if the component is not in the parent type.  */
+  for (cmp = dt->components; cmp; cmp = cmp->next)
+    if (strcmp (c->name, cmp->name) == 0)
+      return;
+
   /* Build a gfc_ref to recursively call gfc_conv_component_ref.  */
   parent.type = REF_COMPONENT;
   parent.next = NULL;
@@ -547,23 +552,11 @@  conv_parent_component_references (gfc_se * se, gfc_ref * ref)
   if (dt->backend_decl == NULL)
     gfc_get_derived_type (dt);
 
-  if (dt->attr.extension && dt->components)
-    {
-      if (dt->attr.is_class)
-	cmp = dt->components;
-      else
-	cmp = dt->components->next;
-      /* Return if the component is not in the parent type.  */
-      for (; cmp; cmp = cmp->next)
-	if (strcmp (c->name, cmp->name) == 0)
-	  return;
-	
-      /* Otherwise build the reference and call self.  */
-      gfc_conv_component_ref (se, &parent);
-      parent.u.c.sym = dt->components->ts.u.derived;
-      parent.u.c.component = c;
-      conv_parent_component_references (se, &parent);
-    }
+  /* Build the reference and call self.  */
+  gfc_conv_component_ref (se, &parent);
+  parent.u.c.sym = dt->components->ts.u.derived;
+  parent.u.c.component = c;
+  conv_parent_component_references (se, &parent);
 }
 
 /* Return the contents of a variable. Also handles reference/pointer