[Bug fortran/49586] New: Multiple initialization with DATA: Warning and initialization order

[Bug fortran/49586] New: Multiple initialization with DATA: Warning and initialization order

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49586

           Summary: Multiple initialization with DATA: Warning and
                    initialization order
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Keywords: diagnostic
          Severity: normal
          Priority: P3
         Component: fortran
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: burnus@gcc.gnu.org
        Depends on: 49540


Follow up to PR 49540.

gfortran - as some other compilers - allows multiple initialization in DATA
statements.

Expected:
- Warn - at least with -Wall - if multiple initialization occur
  (Currently, only -pedantic warns and -std=f* prints an error)
- Check whether the current initialization makes sense
  I think the version of gfortran 4.1 + the commercial compilers makes most
  sense.


For the test case below, the result is (* = warns by default):

- g95, NAG: Error because of multiple initialization

- gfortran 4.1 (!), ifort, sunf90, pathf90*, openf90*, pgf90:
4, 5, 5, 5, 5,  5, 5, 5, 5, 5,  5, 5, 5, 5, 5,  5, 5, 5, 5, 5,  5, 5, 5, 5, 6

- gfortran 4.6 and 4.7:
4, 5, 5, 5, 5,  1, 2, 2, 2, 3,  5, 5, 5, 5, 5,  5, 5, 5, 5, 5,  5, 5, 5, 5, 6

- gfortran 4.3, 4.4 and 4.5:
4, 3, 3, 3, 3,  3, 3, 3, 3, 3,  3, 3, 3, 3, 3,  3, 3, 3, 3, 3,  3, 3, 3, 3, 6


      program pr49540_2
        common /a/ i(5,5)
        data i(1:5,2) /1, 3 * 2, 3/
        data i /4, 23 * 5, 6/
        print '(5(5(i0:", "):" "))', i
      end

[Bug fortran/49586] Multiple initialization with DATA: Warning and initialization order

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49586

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-06-30 07:01:01 UTC ---
More complete testcase, which should be adjusted to do runtime checking of the
variable content based on what we decide, and add whatever options will be
needed to disable the errors when compiling it.

block data
  common /a/ i(5,5)
  data i /4, 23 * 5, 6/
  data i(:,2) /1, 3 * 2, 3/
  common /b/ j(5,5)
  data j(2,:) /1, 3 * 2, 3/
  data j /4, 23 * 5, 6/
  common /c/ k(5,5)
  data k(:,2) /1, 3 * 2, 3/
  data k /4, 23 * 5, 6/
  common /d/ l(5,5)
  data l /4, 23 * 5, 6/
  data l(2,:) /1, 3 * 2, 3/
end block data

[Bug fortran/49586] Multiple initialization with DATA: Warning and initialization order

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49586

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-06-30 11:43:24 UTC ---
The order apparently changed with
http://gcc.gnu.org/viewcvs?view=revision&revision=114406
which doesn't change the order actually, but only wouldn't overwrite anything.
So, even back then the DATA stmts used to be traversed in wrong order.

[Bug fortran/49586] Multiple initialization with DATA: Warning and initialization order

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49586

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-06-30 12:30:51 UTC ---
Created attachment 24645
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=24645
gcc47-pr49586.patch

Untested fix for the order.

[Bug fortran/49586] Multiple initialization with DATA: Warning and initialization order

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49586

--- Comment #4 from Tobias Burnus <burnus at gcc dot gnu.org> 2011-06-30 13:57:21 UTC ---
(In reply to comment #3)
> Untested fix for the order.

For the example of comment 0, gfortran + patch gives the gfortran 4.1 et al.
result.

For the example which is contained in the attachment, one gets the following
result, when printing (in that order) i, j, k, l.

gfortran 4.1-4.5: ICE in gfc_assign_data_value_range.

gfortran 4.7 (w/ patch) -- and pathf90 3.2.99, openf95 4.2 and sunf95 8.3:
4, 5, 5, 5, 5, 1, 2, 2, 2, 3, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6
4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6
4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6
4, 1, 5, 5, 5, 5, 2, 5, 5, 5, 5, 2, 5, 5, 5, 5, 2, 5, 5, 5, 5, 3, 5, 5, 6

ifort 11.1 and pgf90 10.1:
4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6
4, 1, 0, 0, 0, 0, 2, 0, 0, 0, 0, 2, 0, 0, 0, 0, 2, 0, 0, 0, 0, 3, 0, 0, 6
4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6
4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6

gfortran 4.6/4.7 (w/o patch):
4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6
4, 1, 5, 5, 5, 5, 2, 5, 5, 5, 5, 2, 5, 5, 5, 5, 2, 5, 5, 5, 5, 3, 5, 5, 6
4, 5, 5, 5, 5, 1, 2, 2, 2, 3, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6
4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6

For comparison, the m and n lines are:
4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6
4, 5, 5, 5, 5, 1, 2, 2, 2, 3, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6

[Bug fortran/49586] Multiple initialization with DATA: Warning and initialization order

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49586

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-06-30 16:10:18 UTC ---
Unfortunately, this patch breaks gfortran.fortran-torture/execute/data.f90
compilation, it fails with a cryptic error.
No idea if it is a bug in the testcase that just relied on the older ordering,
or if my patch is buggy.  I guess I give up here.

[Bug fortran/49586] Multiple initialization with DATA: Warning and initialization order

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49586

--- Comment #6 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2011-06-30 17:28:48 UTC ---
> Unfortunately, this patch breaks gfortran.fortran-torture/execute/data.f90
> compilation, it fails with a cryptic error.

Confirmed. A reduced test is

        ! Program to test data statement
        program data
        call sub1()
        end
        subroutine sub1()
        integer i
        type tmp
          integer, dimension(4)::a 
          real :: r 
        end type
        type tmp1
          type (tmp) t1(4)
          integer b
        end type
        type (tmp1) tmp2(2)
        ! Full array and scalar component initializer
        data tmp2(2)%t1(2)%r, tmp2(1)%t1(3)%a,
tmp2(1)%b/220,136,137,138,139,10/
        data tmp2(2)%t1(2)%a(3), tmp2(2)%t1(2)%a(1)/223,221/

        end

The error is emitted in show_loci in gcc/fortran/error.c when
l1 == NULL || l1->lb == NULL

With gdb I get

Breakpoint 1, show_loci (l1=0x141d23260, l2=0x0) at
../../work/gcc/fortran/error.c:383
383      if (l1 == NULL || l1->lb == NULL)
(gdb) p l1
$1 = (locus *) 0x141d23260
(gdb) p *l1
$2 = {
  nextc = 0x0, 
  lb = 0x0
}
(gdb) bt
#0  show_loci (l1=0x141d23260, l2=0x0) at ../../work/gcc/fortran/error.c:383
#1  0x000000010002832d in error_print (type=0x100a46df1 "Error:",
format0=0x100a63ad0 "non-constant initialization expression at %L",
argp=0x7fff5fbfd128) at ../../work/gcc/fortran/error.c:627
Previous frame inner to this frame (gdb could not unwind past this frame)

[Bug fortran/49586] Multiple initialization with DATA: Warning and initialization order

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49586

Dominique d'Humieres <dominiq at lps dot ens.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |DUPLICATE

--- Comment #7 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
Duplicate of pr45288.

*** This bug has been marked as a duplicate of bug 45288 ***

[Bug fortran/49586] Multiple initialization with DATA: Warning and initialization order

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49586

Dominique d'Humieres <dominiq at lps dot ens.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |DUPLICATE
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |DUPLICATE

--- Comment #7 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
Duplicate of pr45288.

*** This bug has been marked as a duplicate of bug 45288 ***

--- Comment #8 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
Duplicate of pr45288.

*** This bug has been marked as a duplicate of bug 45288 ***