[Bug fortran/50790] New: ICE in copy_constant on divide by expression that evaluates to zero

[Bug fortran/50790] New: ICE in copy_constant on divide by expression that evaluates to zero

[jonathan.hogg at stfc dot ac.uk]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50790

             Bug #: 50790
           Summary: ICE in copy_constant on divide by expression that
                    evaluates to zero
    Classification: Unclassified
           Product: gcc
           Version: 4.6.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: fortran
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: jonathan.hogg@stfc.ac.uk


The following code produces an ICE for me:

program test
   implicit none

   integer :: i

   i = 3

   print *, "chk", 3/(i-i)
end program test

This should presumably produce
   Error: Division by zero
at compile time (or run time if we're not being clever).

Produced with:
GNU Fortran (Debian 4.6.1-4) 4.6.1
Copyright (C) 2011 Free Software Foundation, Inc.

Thanks,
Jonathan.

[Bug middle-end/50790] ICE in copy_constant on divide by expression that evaluates to zero

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50790

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |burnus at gcc dot gnu.org
          Component|fortran                     |middle-end

--- Comment #1 from Tobias Burnus <burnus at gcc dot gnu.org> 2011-10-19 15:23:55 UTC ---
-fdump-tree-original:

      static integer(kind=4) C.1534 = 3 / 0;
      _gfortran_transfer_integer_write (&dt_parm.0, &C.1534, 4);

It gives with 4.7 the ICE:
  internal compiler error: in copy_constant, at varasm.c:3031

I think it is - at least in general - not be detectable in the front end but
only in the ME.



A similar C test case works (cf. below for a simpler version); I think there is
no compiler option to print a warning -- at least -Wall -Wextra don't.

int
main ()
{
  int i = 3;
  printf ("%d\n", 3/(i-i));
  return 0;
}

[Bug middle-end/50790] ICE in copy_constant on divide by expression that evaluates to zero

[kargl at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50790

kargl at gcc dot gnu.org changed:

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

--- Comment #2 from kargl at gcc dot gnu.org 2011-10-19 17:54:31 UTC ---
Here's a variation on the original code, which removes
the print statement.  One does not need to wade through
gfortran IO structure.

program test
   implicit none
   integer i
   i = 3
   call foo (3 / (i - i))
end program test

subroutine foo(j)
  j = j * 1
end subroutine foo

The dump shows

foo (integer(kind=4) & restrict j)
{
  *j = NON_LVALUE_EXPR <*j>;
}


test ()
{
  integer(kind=4) i;

  i = 3;
  {
    static integer(kind=4) C.1729 = 3 / 0;

    foo (&C.1729);
  }
}


An equivalent C program compiles and gives a nearly identical
dump, but dies with a Floating exception(?)

troutmask:sgk[247] cat dw.c
void
foo (int *j)
{
  *j = *j * 2;
}

int
main()
{
  static int i = 3;
  i = 3 / (i - i);
  foo(&i);
  return 0;
}
troutmask:sgk[248] ~/work/4x/bin/gcc -o z dw.c && ./z
Floating exception (core dumped)

[Bug middle-end/50790] ICE in copy_constant on divide by expression that evaluates to zero

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=50790

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
   Target Milestone|---                         |8.0
      Known to work|                            |8.1.0
      Known to fail|                            |4.9.4, 5.5.0, 6.3.0, 7.1.0,
                   |                            |7.2.0, 7.3.0
             Status|WAITING                     |RESOLVED
           Keywords|                            |ice-on-valid-code

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Fixed.