Re: Behaviour with -fcheck=all

Re: Behaviour with -fcheck=all

[Dominique d'Humières]

Hi Salvatore,

>  Consider the attached code. …

This is pr52162. If you want it for older revisions you can try to apply the patch for trans-expr.c at revisions r256607.

Cheers,

Dominique

Behaviour with -fcheck=all

[Salvatore Filippone]

[-- Attachment #1: Type: text/plain, Size: 3031 bytes --]

Hello everybody
Consider the attached code. At line 11 we do have both allocation on
assignment and KIND conversion; my reading of the standard is that
this is allowed. And indeed if I compile with "normal" options I do
get the expected output.
However, if I compile with -fcheck=all, the runtime throws me an error
with 6.3 and 7.3, but works with 8.0. I did not find anything obivious
in bugzilla, the closest thing I can see is PR67430 but there is no
intrinsic vs defined assignment here.
Feel free to add to the test suite if appropriate; in the meantime
I'll be forced to do some defensive programming, since so much of my
code has to run on platforms down to GCC version 4.9, and I suppose
that even in the best circumstances, whatever fixed this in 8 will not
be backported that far.

Cheers
Salvatore


====================================
[sfilippo@localhost NewMixedI8]$ gfortran -v
Using built-in specs.
COLLECT_GCC=gfortran
COLLECT_LTO_WRAPPER=/opt/gnu/7.3.0/libexec/gcc/x86_64-pc-linux-gnu/7.3.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../gcc-7.3.0/configure --prefix=/opt/gnu/7.3.0
--enable-languages=c,c++,fortran
Thread model: posix
gcc version 7.3.0 (GCC)
[sfilippo@localhost NewMixedI8]$ gfortran -o try_vassign try_vassign.f90
[sfilippo@localhost NewMixedI8]$ ./try_vassign
                    5                    6                    7
            8
                   -4                    6                    7
            8
[sfilippo@localhost NewMixedI8]$ gfortran -o try_vassign
try_vassign.f90 -ggdb -O0 -fcheck=all -fbacktrace
[sfilippo@localhost NewMixedI8]$ ./try_vassign
                    5                    6                    7
            8
At line 11 of file try_vassign.f90
Fortran runtime error: Array bound mismatch for dimension 1 of array 'w' (0/4)

Error termination. Backtrace:
#0  0x400b35 in __foobar_mod_MOD_foo
    at /home/sfilippo/NUMERICAL/PSBLAS/GitHub/NewMixedI8/try_vassign.f90:11
#1  0x4013c6 in try
    at /home/sfilippo/NUMERICAL/PSBLAS/GitHub/NewMixedI8/try_vassign.f90:28
#2  0x401474 in main
    at /home/sfilippo/NUMERICAL/PSBLAS/GitHub/NewMixedI8/try_vassign.f90:22

[sfilippo@localhost NewMixedI8]$ module load gnu/8.0.0
    gnu/8.0.0 - loads the GNU 8.0.0 compilers suite

    Version 8.0.0

[sfilippo@localhost NewMixedI8]$ gfortran -v
Using built-in specs.
COLLECT_GCC=gfortran
COLLECT_LTO_WRAPPER=/opt/gnu/8.0.0/libexec/gcc/x86_64-pc-linux-gnu/8.0.1/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../gcc/configure --prefix=/opt/gnu/8.0.0
--enable-languages=c,c++,fortran,lto --no-create --no-recursion
Thread model: posix
gcc version 8.0.1 20180228 (experimental) (GCC)
[sfilippo@localhost NewMixedI8]$ gfortran -o try_vassign
try_vassign.f90 -ggdb -O0 -fcheck=all -fbacktrace
[sfilippo@localhost NewMixedI8]$ ./try_vassign
                    5                    6                    7
            8
                   -4                    6                    7
            8

==================================================

[-- Attachment #2: try_vassign.f90 --]
[-- Type: text/x-fortran, Size: 543 bytes --]

module foobar_mod
contains
  subroutine foo(x)
    use iso_fortran_env
    implicit none
    integer(int64), allocatable, intent(inout)   :: x(:)
    
    integer(int32), allocatable  :: w(:)
    
    if (allocated(x)) then
      w = x
      if (size(w) > 0) then 
        w(1) = 1 - w(1)
      end if
      x = w
    end if
  end subroutine foo
end module foobar_mod
  
program try
  use iso_fortran_env
  use foobar_mod

  integer(int64), allocatable :: v(:)

  v = (/5,6,7,8/)
  write(*,*) v
  call foo(v)
  write(*,*) v
end program try