[Bug fortran/46818] New: ICE on pointer assignment

[Bug fortran/46818] New: ICE on pointer assignment

[m.a.hulsen at tue dot nl]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46818

           Summary: ICE on pointer assignment
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: blocker
          Priority: P3
         Component: fortran
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: m.a.hulsen@tue.nl


Created attachment 22653
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=22653
all the files

After typing 
make
I get:
gfortran -c  -O2    kind_defs.f90
gfortran -c  -O2    system_defs.f90
gfortran -c  -O2    system_vector.f90
gfortran -c  -O2    system.f90
gfortran -c  -O2    t.f90
t.f90: In function ‘t’:
t.f90:20:0: internal compiler error: in fold_convert_loc, at fold-const.c:2028
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.
make: *** [t.o] Error 1
[martien@pc379 tmp]$ gfortran --version
GNU Fortran (GCC) 4.6.0 20101206 (experimental) [trunk revision 167484]
Copyright (C) 2010 Free Software Foundation, Inc.

Note, that if I concatenate all files into a single one, the ICE is gone.

[Bug fortran/46818] [4.6 Regression] ICE on pointer assignment

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46818

Tobias Burnus <burnus at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
      Known to work|                            |4.3.4
           Keywords|                            |ice-on-valid-code
   Last reconfirmed|                            |2010.12.06 13:05:28
                 CC|                            |burnus at gcc dot gnu.org
     Ever Confirmed|0                           |1
            Summary|ICE on pointer assignment   |[4.6 Regression] ICE on
                   |                            |pointer assignment
   Target Milestone|---                         |4.6.0
      Known to fail|                            |4.6.0
           Severity|blocker                     |normal

--- Comment #1 from Tobias Burnus <burnus at gcc dot gnu.org> 2010-12-06 13:05:28 UTC ---
Confirm. Works with GCC 4.3.4, fails with GCC 4.6.0.

TODO: Check whether it fails also on 4.4 or 4.5 (and update summary/target
milestone accordingly).


Reduced test case:

! ============== system_defs.f90 =============
module system_defs_m
  type sysvector_t
    integer :: probnr = 0
    real, allocatable, dimension(:) :: u
  end type sysvector_t
end module system_defs_m

! ========== t.f90 ===========================
module convecreac_m
  use system_defs_m
  type(sysvector_t), pointer :: solution
end module convecreac_m

program t
  use convecreac_m
  implicit none
  type(sysvector_t), target :: sol
  solution => sol
end program t

[Bug fortran/46818] [4.6 Regression] ICE on pointer assignment

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46818

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P4
      Known to work|                            |4.4.4, 4.5.0, 4.5.1

[Bug fortran/46818] [4.6 Regression] ICE on pointer assignment

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46818

--- Comment #2 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2010-12-06 13:26:42 UTC ---
The test in comment #1 compiles with  -fno-whole-file, so the regression comes
from -fwhole-file being the default (compiles with r162456, ICE with r166102).

[Bug fortran/46818] [4.6 Regression] ICE on pointer assignment (-fwhole-file)

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46818

Tobias Burnus <burnus at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Depends on|                            |44945
            Summary|[4.6 Regression] ICE on     |[4.6 Regression] ICE on
                   |pointer assignment          |pointer assignment
                   |                            |(-fwhole-file)

--- Comment #3 from Tobias Burnus <burnus at gcc dot gnu.org> 2010-12-06 13:39:30 UTC ---
Well, then it is probably in a way a duplicate of PR 44945 => Mark this PR as
depending on PR 44945.

[Bug fortran/46818] [4.6 Regression] ICE on pointer assignment (-fwhole-file)

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46818

--- Comment #4 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2010-12-06 13:49:25 UTC ---
> Well, then it is probably in a way a duplicate of PR 44945 => Mark this PR as
> depending on PR 44945.

I don't think it is a duplicate of  PR 44945, the failure is the same than with
the original code of pr40440 (which has always given an ICE when the
iso_varying_string was not included in the file, see also
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40011#c57 ).

[Bug fortran/46818] [4.6 Regression] ICE on pointer assignment (-fwhole-file)

[pault at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46818

Paul Thomas <pault at gcc dot gnu.org> changed:

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

--- Comment #5 from Paul Thomas <pault at gcc dot gnu.org> 2010-12-27 18:10:51 UTC ---
(In reply to comment #4)

Separating

module convecreac_m

  use kind_defs_m
  use system_m, only: sysvector_t

  implicit none

  type(sysvector_t), pointer :: solution

end module convecreac_m

into a new file (equivalent to -fno-whole-file) also clears the problem.  It
seems that we have another TYPE_CANONICAL problem.

Paul

[Bug fortran/46818] [4.6 Regression] ICE on pointer assignment (-fwhole-file)

[burnus@net-b.de]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46818

--- Comment #6 from burnus@net-b.de <burnus@net-b.de> 2010-12-29 10:55:11 UTC ---
(In reply to comment #5)
> It seems that we have another TYPE_CANONICAL problem.

Kind of. The problem (for this PR and for PR 46818) is that one only 
generates a gsym for a MODULE; one should also generate a gsym if one 
USEs a symbol from a module for which no gsym exists. Currently, one 
generates multiple decls if one USE-associates the same symbole multiple 
times.

[Bug fortran/46818] [4.6 Regression] ICE on pointer assignment (-fwhole-file)

[pault at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46818

--- Comment #7 from Paul Thomas <pault at gcc dot gnu.org> 2011-02-20 17:00:50 UTC ---
Author: pault
Date: Sun Feb 20 17:00:47 2011
New Revision: 170338

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=170338
Log:
2011-02-20  Paul Thomas  <pault@gcc.gnu.org>

    PR fortran/46818
    * gfortran.dg/whole_file_30.f90 : New test.
    * gfortran.dg/whole_file_31.f90 : New test.


Added:
    trunk/gcc/testsuite/gfortran.dg/whole_file_30.f90
    trunk/gcc/testsuite/gfortran.dg/whole_file_31.f90
Modified:
    trunk/gcc/testsuite/ChangeLog

[Bug fortran/46818] [4.6 Regression] ICE on pointer assignment (-fwhole-file)

[pault at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46818

Paul Thomas <pault at gcc dot gnu.org> changed:

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

--- Comment #8 from Paul Thomas <pault at gcc dot gnu.org> 2011-02-20 17:03:43 UTC ---
This was in fact put right by the fix for pr45077.  However, I have added the
reduced testcase of comment #1.

Fixed on trunk.

Thanks for the report Martien!

Paul