[Bug fortran/35721] New: ASSOCIATED returns false when strides confusing

[Bug fortran/35721] New: ASSOCIATED returns false when strides confusing

[dick dot hendrickson at gmail dot com]

The ASSOCIATED functions returns false for the 4th test below.  It
should return true.  there have been a ton of interps and rewording
of the associated function.  It's probably clearer to read
case (v) in the F2003 standard.

      program try_mg0028
! fails on Windows XP
! gcc version 4.4.0 20080312 (experimental) [trunk revision 133139]

      real  tda2r(2,3)

      call       mg0028(tda2r,  1,  2,  3)

      end


      SUBROUTINE MG0028(TDA2R,nf1,nf2,nf3)
      real, pointer  ::  TLA2L(:,:),TLA2L1(:,:)
      real, target   ::  TDA2R(NF2,NF3)
      logical LL(4)
      TLA2L => TDA2R(NF2:NF1:-NF2,NF3:NF1:-NF2)
      TLA2L1 => TLA2L
      LL(1) = ASSOCIATED(TLA2L)
      LL(2) = ASSOCIATED(TLA2L,TLA2L1)
      LL(3) = ASSOCIATED(TLA2L,TDA2R)
      LL(4) = ASSOCIATED(TLA2L1,TDA2R(2:2,3:1:-2))  !should be true

      if (any(LL .neqv. (/ .true., .true., .false., .true./))) then
        print *, LL
        print *, shape(TLA2L1)
        print *, shape(TDA2R(2:2,3:1:-2))
      endif

      END SUBROUTINE


C:\g_experiments\gfortran>gfortran m
g0028.f

C:\g_experiments\gfortran>a
 T T F F
           1           2
           1           2


-- 
           Summary: ASSOCIATED returns false when strides confusing
           Product: gcc
           Version: 4.4.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: fortran
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: dick dot hendrickson at gmail dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35721

[Bug fortran/35721] ASSOCIATED returns false when strides confusing

[burnus at gcc dot gnu dot org]

------- Comment #1 from burnus at gcc dot gnu dot org  2008-03-27 18:27 -------
Confirm. Note NAG f95 complains the program is invalid and I think it is right:

Error: Explicit interface required for MG0028 from TRY_MG0028 - argument TDA2R
(no. 1) is a TARGET

But this does not solve the gfortran problem ;-)

>From the standard: (v) applies here, but (vii) is analog - except that TARGET
is a pointer instead of a target.

"Case (v): If TARGET is present and is an array target, the result is true if
the target associated with POINTER and TARGET have the same shape, are neither
of size zero nor arrays whose elements are zero-sized storage sequences, and
occupy the same storage units in array element order. Otherwise, the result is
false. If POINTER is disassociated, the result is false."

gfortran is failing since in dimension 1 they have different strides:
Pointer:  2:1:-2  (namely: element 2)
Target:   2:2:1   (namely: element 2)

However, libgfortran/intrinsics/associated.c only checks whether the stride is
the same. Proposed patch:

Index: libgfortran/intrinsics/associated.c
===================================================================
--- libgfortran/intrinsics/associated.c (Revision 133633)
+++ libgfortran/intrinsics/associated.c (Arbeitskopie)
@@ -48,10 +48,12 @@ associated (const gfc_array_void *pointe
   rank = GFC_DESCRIPTOR_RANK (pointer);
   for (n = 0; n < rank; n++)
     {
-      if (pointer->dim[n].stride != target->dim[n].stride)
+      long extent;
+      extent = pointer->dim[n].ubound - pointer->dim[n].lbound;
+
+      if (extent != (target->dim[n].ubound - target->dim[n].lbound))
         return 0;
-      if ((pointer->dim[n].ubound - pointer->dim[n].lbound)
-          != (target->dim[n].ubound - target->dim[n].lbound))
+      if (pointer->dim[n].stride != target->dim[n].stride && extent != 0)
         return 0;
       if (pointer->dim[n].ubound < pointer->dim[n].lbound)
        return 0;


-- 

burnus at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |burnus at gcc dot gnu dot
                   |                            |org
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2008-03-27 18:27:43
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35721

[Bug fortran/35721] ASSOCIATED returns false when strides confusing

[dick dot hendrickson at gmail dot com]

------- Comment #2 from dick dot hendrickson at gmail dot com  2008-03-28 00:57 -------
Subject: Re:  ASSOCIATED returns false when strides confusing

On 27 Mar 2008 18:27:44 -0000, burnus at gcc dot gnu dot org
<gcc-bugzilla@gcc.gnu.org> wrote:
>
>
>  ------- Comment #1 from burnus at gcc dot gnu dot org  2008-03-27 18:27 -------
>  Confirm. Note NAG f95 complains the program is invalid and I think it is right:
>
>  Error: Explicit interface required for MG0028 from TRY_MG0028 - argument TDA2R
>  (no. 1) is a TARGET
>
Yes, you (and NAG) are correct.  Originally, the subroutine was in a big module
and when I cut it out I did the quick and dirty thing of merely
calling it without
thinking.  For what it's worth, the other g compiler I tried this on
also got the wrong
answer and didn't complain about the interface.

Dick Hendrickson

>  But this does not solve the gfortran problem ;-)
>
>  From the standard: (v) applies here, but (vii) is analog - except that TARGET
>  is a pointer instead of a target.
>
>  "Case (v): If TARGET is present and is an array target, the result is true if
>  the target associated with POINTER and TARGET have the same shape, are neither
>  of size zero nor arrays whose elements are zero-sized storage sequences, and
>  occupy the same storage units in array element order. Otherwise, the result is
>  false. If POINTER is disassociated, the result is false."
>
>  gfortran is failing since in dimension 1 they have different strides:
>  Pointer:  2:1:-2  (namely: element 2)
>  Target:   2:2:1   (namely: element 2)
>
>  However, libgfortran/intrinsics/associated.c only checks whether the stride is
>  the same. Proposed patch:
>
>  Index: libgfortran/intrinsics/associated.c
>  ===================================================================
>  --- libgfortran/intrinsics/associated.c (Revision 133633)
>  +++ libgfortran/intrinsics/associated.c (Arbeitskopie)
>  @@ -48,10 +48,12 @@ associated (const gfc_array_void *pointe
>    rank = GFC_DESCRIPTOR_RANK (pointer);
>    for (n = 0; n < rank; n++)
>      {
>  -      if (pointer->dim[n].stride != target->dim[n].stride)
>  +      long extent;
>  +      extent = pointer->dim[n].ubound - pointer->dim[n].lbound;
>  +
>  +      if (extent != (target->dim[n].ubound - target->dim[n].lbound))
>          return 0;
>  -      if ((pointer->dim[n].ubound - pointer->dim[n].lbound)
>  -          != (target->dim[n].ubound - target->dim[n].lbound))
>  +      if (pointer->dim[n].stride != target->dim[n].stride && extent != 0)
>          return 0;
>        if (pointer->dim[n].ubound < pointer->dim[n].lbound)
>         return 0;
>
>
>  --
>
>  burnus at gcc dot gnu dot org changed:
>
>            What    |Removed                     |Added
>  ----------------------------------------------------------------------------
>                  CC|                            |burnus at gcc dot gnu dot
>                    |                            |org
>              Status|UNCONFIRMED                 |NEW
>      Ever Confirmed|0                           |1
>    Last reconfirmed|0000-00-00 00:00:00         |2008-03-27 18:27:43
>                date|                            |
>
>
>  http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35721
>
>  ------- You are receiving this mail because: -------
>  You reported the bug, or are watching the reporter.
>


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35721

[Bug fortran/35721] ASSOCIATED returns false when strides confusing

[burnus at gcc dot gnu dot org]

------- Comment #3 from burnus at gcc dot gnu dot org  2008-03-28 13:47 -------
Subject: Bug 35721

Author: burnus
Date: Fri Mar 28 13:47:06 2008
New Revision: 133684

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=133684
Log:
2008-03-28  Tobias Burnus  <burnus@net-b.de>

        PR fortran/35721
        * intrinsics/associated.c (associated): Ignore different
        stride of pointer vs. target if only one element is referred.

2008-03-28  Tobias Burnus  <burnus@net-b.de>

        PR fortran/35721
        * gfortran.dg/associated_target_2.f90: New.


Added:
    trunk/gcc/testsuite/gfortran.dg/associated_target_2.f90
Modified:
    trunk/gcc/testsuite/ChangeLog
    trunk/libgfortran/ChangeLog
    trunk/libgfortran/intrinsics/associated.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35721

[Bug fortran/35721] ASSOCIATED returns false when strides confusing

[burnus at gcc dot gnu dot org]

------- Comment #4 from burnus at gcc dot gnu dot org  2008-03-28 13:49 -------
FIXED on the trunk (4.4.0). Thanks again for the report.


-- 

burnus at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
           Keywords|                            |wrong-code
         Resolution|                            |FIXED
   Target Milestone|---                         |4.4.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35721