From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-bugs-return-369541-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 16595 invoked by alias); 5 Oct 2011 21:40:38 -0000
Received: (qmail 16582 invoked by uid 22791); 5 Oct 2011 21:40:35 -0000
X-SWARE-Spam-Status: No, hits=-2.9 required=5.0	tests=ALL_TRUSTED,AWL,BAYES_00
X-Spam-Check-By: sourceware.org
Received: from localhost (HELO gcc.gnu.org) (127.0.0.1)    by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Wed, 05 Oct 2011 21:40:22 +0000
From: "dominiq at lps dot ens.fr" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug fortran/47844] Array stride ignored for pointer-valued function results
Date: Wed, 05 Oct 2011 21:40:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: fortran
X-Bugzilla-Keywords: wrong-code
X-Bugzilla-Severity: minor
X-Bugzilla-Who: dominiq at lps dot ens.fr
X-Bugzilla-Status: NEW
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Changed-Fields:
Message-ID: <bug-47844-4-iBqrK6iLa3@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-47844-4@http.gcc.gnu.org/bugzilla/>
References: <bug-47844-4@http.gcc.gnu.org/bugzilla/>
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
Content-Type: text/plain; charset="UTF-8"
MIME-Version: 1.0
Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-bugs.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-help@gcc.gnu.org>
Sender: gcc-bugs-owner@gcc.gnu.org
X-SW-Source: 2011-10/txt/msg00285.txt.bz2

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47844

--- Comment #8 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2011-10-05 21:39:31 UTC ---
> We could fix this in 4.7 by adding a sm field to array descriptors.

I don't see why. I have looked at the dump.original of the following code:

  integer, target :: tgt(9) = [1,2,3,4,5,6,7,8,9]
  integer, pointer :: p(:)
  integer :: var(3)
  p => tgt(ubound(tgt,1)-4:lbound(tgt,1):-2)
  var = p
  print *, var ! prints 5 3 1
  var = f(tgt) ! should assign 5 3 1
  print *, var ! but    prints 5 6 7
  if(any(var/=[5,3,1])) call abort()
contains
  function f(x) result(r)
    integer, target :: x(:)
    integer, pointer :: r(:)
    r => x(ubound(x,1)-4:lbound(x,1):-2)
    print *, r ! prints 5 3 1
  end function f
end

The first assignment (var = p) is

  p.dtype = 265;
  p.dim[0].lbound = 1;
  p.dim[0].ubound = 3;
  p.dim[0].stride = -2;
  p.data = (void *) &tgt[4];
  p.offset = 2;
  {
    integer(kind=8) D.1793;
    integer(kind=8) D.1792;
    integer(kind=8) D.1791;
    integer(kind=8) D.1790;
    integer(kind=4)[0:] * D.1789;

    D.1789 = (integer(kind=4)[0:] *) p.data;
    D.1790 = p.offset;
    D.1791 = p.dim[0].lbound;
    D.1792 = p.dim[0].ubound;
    D.1793 = D.1791 + -1;
    {
      integer(kind=8) D.1795;
      integer(kind=8) S.7;

      D.1795 = p.dim[0].stride;
      S.7 = 1;
      while (1)
        {
          if (S.7 > 3) goto L.1;
          var[S.7 + -1] = (*D.1789)[(S.7 + D.1793) * D.1795 + D.1790];
          S.7 = S.7 + 1;
        }
      L.1:;
    }
  }

while the second one (var = f(tgt)) is:

    f (&atmp.12, D.1804);
    {
      integer(kind=8) S.13;

      S.13 = 0;
      while (1)
        {
          if (S.13 > 2) goto L.2;
          var[S.13] = (*(integer(kind=4)[3] * restrict) atmp.12.data)[S.13];
          S.13 = S.13 + 1;
        }
      L.2:;
    }

As far I understand the dump, atmp.12 is set with the same values as p (through
an awfully complicated process), but atmp.12.dim[0].stride is not used in the
assignment.