[Bug fortran/97272] New: Wrong answer from MAXLOC with character arg

[Bug fortran/97272] New: Wrong answer from MAXLOC with character arg

[longb at cray dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97272

            Bug ID: 97272
           Summary: Wrong answer from MAXLOC with character arg
           Product: gcc
           Version: 10.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: fortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: longb at cray dot com
  Target Milestone: ---

Test case:

> cat test.f90
      program test
      character, allocatable :: a(:)
      integer(8) l, i
      l = 2000000_8
      allocate (a(l))
      do i = 1, l
      a(i) = 'a'
      end do
      l = l - 1
      a(l) = 'b'
      i = maxloc (a, dim=1, kind=8)
      print *, 'i = ', i, 'a(i) = ', a(i)
      end

Expected behavior (Cray):

> ftn test.f90
> ./a.out
 i =  1999999 a(i) = b

Incorrect result (gfortran):

> module swap PrgEnv-cray PrgEnv-gnu
> gfortran test.f90
> ./a.out
 i =               2000000 a(i) = a
> gfortran --version
GNU Fortran (GCC) 10.1.0 20200507 (Cray Inc.)
Copyright (C) 2020 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

[Bug fortran/97272] Wrong answer from MAXLOC with character arg

[anlauf at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97272

anlauf at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2020-10-02
           Keywords|                            |wrong-code
                 CC|                            |anlauf at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #1 from anlauf at gcc dot gnu.org ---
Confirmed for the case where the kind argument is present:

      i = maxloc (a, dim=1, kind=8)

or

      i = maxloc (a, dim=1, kind=4)

The issue does not occur when the kind argument is not present:

      i = maxloc (a, dim=1)

[Bug fortran/97272] Wrong answer from MAXLOC with character arg

[anlauf at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97272

--- Comment #2 from anlauf at gcc dot gnu.org ---
Untested fix:

diff --git a/gcc/fortran/trans-intrinsic.c b/gcc/fortran/trans-intrinsic.c
index 3b3bd8629cd..9e9898c2bbf 100644
--- a/gcc/fortran/trans-intrinsic.c
+++ b/gcc/fortran/trans-intrinsic.c
@@ -5211,7 +5211,9 @@ gfc_conv_intrinsic_minmaxloc (gfc_se * se, gfc_expr *
expr, enum tree_code op)
       while (a->next)
        {
          b = a->next;
-         if (b->expr == NULL || strcmp (b->name, "dim") == 0)
+         if (b->expr == NULL
+             || strcmp (b->name, "dim") == 0
+             || strcmp (b->name, "kind") == 0)
            {
              a->next = b->next;
              b->next = NULL;

[Bug fortran/97272] Wrong answer from MAXLOC with character arg

[anlauf at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97272

anlauf at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |anlauf at gcc dot gnu.org
             Status|NEW                         |ASSIGNED

--- Comment #3 from anlauf at gcc dot gnu.org ---
Patch: https://gcc.gnu.org/pipermail/fortran/2020-October/055143.html

[Bug fortran/97272] Wrong answer from MAXLOC with character arg

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97272

--- Comment #4 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Harald Anlauf <anlauf@gcc.gnu.org>:

https://gcc.gnu.org/g:35d2c6b6e8a7448a84abbf967feeb78a29117014

commit r11-3646-g35d2c6b6e8a7448a84abbf967feeb78a29117014
Author: Harald Anlauf <anlauf@gmx.de>
Date:   Sun Oct 4 20:24:29 2020 +0200

    PR fortran/97272 - Wrong answer from MAXLOC with character arg

    The optional KIND argument to the MINLOC/MAXLOC intrinsic must not be
    passed to the library function, as the kind conversion of the result
    is treated explicitly elsewhere.

    gcc/fortran/ChangeLog:

            PR fortran/97272
            * trans-intrinsic.c (strip_kind_from_actual): Helper function for
            removal of KIND argument.
            (gfc_conv_intrinsic_minmaxloc): Ignore KIND argument here, as it
            is treated elsewhere.

    gcc/testsuite/ChangeLog:

            PR fortran/97272
            * gfortran.dg/pr97272.f90: New test.

[Bug fortran/97272] Wrong answer from MAXLOC with character arg

[longb at cray dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97272

--- Comment #5 from Bill Long <longb at cray dot com> ---
The original intent of adding the KIND argument was because some
implementations used a 32-bit integer for the result, and it is possible for
the answer to be larger than 2**31-1.  Just checking to be sure that the
underlying library code returns a 64-bit integer that can later be converted
based on the KIND value.

[Bug fortran/97272] Wrong answer from MAXLOC with character arg

[anlauf at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97272

--- Comment #6 from anlauf at gcc dot gnu.org ---
(In reply to Bill Long from comment #5)
> The original intent of adding the KIND argument was because some
> implementations used a 32-bit integer for the result, and it is possible for
> the answer to be larger than 2**31-1.  Just checking to be sure that the
> underlying library code returns a 64-bit integer that can later be converted
> based on the KIND value.

Did you try?  It's not exactly fast, but

program test
  implicit none
  integer(8) :: k, l = 10 + 2_8**32
  character, allocatable :: a(:)
  allocate (a(l))
  print *, l
  a(:) = 'a'
  l = l - 1
  a(l) = 'b'
  print *, l
  print *, size (a, kind=8)
  k = maxloc (a, dim=1, kind=8, back=.true.)
 print *, 'k = ', k, 'a(k) = ', a(k)
end

works for me on master now on a 64-bit platform:

           4294967306
           4294967305
           4294967306
 k =            4294967305 a(k) = b

[Bug fortran/97272] Wrong answer from MAXLOC with character arg

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97272

--- Comment #7 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-10 branch has been updated by Harald Anlauf
<anlauf@gcc.gnu.org>:

https://gcc.gnu.org/g:d4ec0a15afb7538247206aa9936071544fd860f8

commit r10-8870-gd4ec0a15afb7538247206aa9936071544fd860f8
Author: Harald Anlauf <anlauf@gmx.de>
Date:   Sun Oct 4 20:24:29 2020 +0200

    PR fortran/97272 - Wrong answer from MAXLOC with character arg

    The optional KIND argument to the MINLOC/MAXLOC intrinsic must not be
    passed to the library function, as the kind conversion of the result
    is treated explicitly elsewhere.

    gcc/fortran/ChangeLog:

            PR fortran/97272
            * trans-intrinsic.c (strip_kind_from_actual): Helper function for
            removal of KIND argument.
            (gfc_conv_intrinsic_minmaxloc): Ignore KIND argument here, as it
            is treated elsewhere.

    gcc/testsuite/ChangeLog:

            PR fortran/97272
            * gfortran.dg/pr97272.f90: New test.

    (cherry picked from commit 35d2c6b6e8a7448a84abbf967feeb78a29117014)

[Bug fortran/97272] Wrong answer from MAXLOC with character arg

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97272

--- Comment #8 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-9 branch has been updated by Harald Anlauf
<anlauf@gcc.gnu.org>:

https://gcc.gnu.org/g:83967508034454425dfce7fe0ea33a153c34e7cb

commit r9-8985-g83967508034454425dfce7fe0ea33a153c34e7cb
Author: Harald Anlauf <anlauf@gmx.de>
Date:   Sun Oct 4 20:24:29 2020 +0200

    PR fortran/97272 - Wrong answer from MAXLOC with character arg

    The optional KIND argument to the MINLOC/MAXLOC intrinsic must not be
    passed to the library function, as the kind conversion of the result
    is treated explicitly elsewhere.

    gcc/fortran/ChangeLog:

            PR fortran/97272
            * trans-intrinsic.c (strip_kind_from_actual): Helper function for
            removal of KIND argument.
            (gfc_conv_intrinsic_minmaxloc): Ignore KIND argument here, as it
            is treated elsewhere.

    gcc/testsuite/ChangeLog:

            PR fortran/97272
            * gfortran.dg/pr97272.f90: New test.

    (cherry picked from commit 35d2c6b6e8a7448a84abbf967feeb78a29117014)

[Bug fortran/97272] Wrong answer from MAXLOC with character arg

[anlauf at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97272

anlauf at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|ASSIGNED                    |RESOLVED

--- Comment #9 from anlauf at gcc dot gnu.org ---
Fixed on master for gcc-11, and on 10- and 9-branch.

Not applied to 8-branch, which did not support the back argument,
so not considered worth it.

Thanks for the report!

[Bug fortran/97272] Wrong answer from MAXLOC with character arg

[longb at cray dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97272

--- Comment #10 from Bill Long <longb at cray dot com> ---
Still fails with 10.2.0.  Can you say which release version will include the
fix?

[Bug fortran/97272] Wrong answer from MAXLOC with character arg

[anlauf at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97272

--- Comment #11 from anlauf at gcc dot gnu.org ---
(In reply to Bill Long from comment #10)
> Still fails with 10.2.0.  Can you say which release version will include the
> fix?

According to https://gcc.gnu.org/, gcc 10.2 was released in July 2020.
Extrapolating previous schedules, I'd expect 10.3 one year after 9.3,
which was released in March 2020.

You could check the 10-branch that the fix works for you.