Bug with IFX?

[Robert_van_Amerongen]

As it becomes time to move from the classic compiler to the new IFX compiler, I start to use the IFX compiler. My first attempt, however, failed when calling C_F_POINTER. The compiler invites me friendly to report this internal compiler error. So here I am.

With the classic compiler I did not see this error for either this example as many other projects where I use it. I have enclosed a small example.

Robert

[Steve_Lionel]

Just a general comment - it is helpful for others to give a more descriptive title to threads here. You could have written something like "ICE with IFX and C_F_POINTER".

[FortranFan]

@Robert_van_Amerongen ,

You can always claim an "internal compiler error" (ICE) is a bug and as pointed to you, mentioning ICE in the thread title itself can help catch the attention of Intel team.

Separately, if your interest is in an "early" implementation of the Fortran 2023 "c_f_strpointer" intrinsic module procedure, you may want to consider simplifying greatly what you are doing, avoid any unnecessary allocations, etc.  Here's a snippet you can try with the new compiler, especially IFX - note I haven't tested it, it's a quick edit on a tab based on your example with my mental compiler:

module cstring_m
   use, intrinsic :: iso_c_binding, only : c_char, c_size_t, c_ptr, c_f_pointer
   generic :: c_f_strpointer => c_f_strpointer_cstrptr
contains
   subroutine c_f_strpointer_cstrptr( cstrptr, fstrptr, nchars )
      ! Argument list
      type(c_ptr), intent(in)                             :: cstrptr
      character(kind=c_char, len=:), pointer, intent(out) :: fstrptr
      integer(c_size_t), intent(in)                       :: nchars
      if ( nchars >= 0 ) then
         block 
            character(kind=c_char, len=nchars), pointer :: sptr
            call c_f_pointer( cstrptr, sptr )
            fstrptr => sptr
            sptr => null()
         end block
      end if
      return
   end subroutine 
end module
   use, intrinsic :: iso_c_binding, only : c_char, c_null_char, c_loc, c_size_t
   use cstring_m, only : c_f_strpointer
   character(kind=c_char, len=:), allocatable, target :: s
   character(kind=c_char, len=:), pointer :: ps
   s = c_char_"This is a test string!" // c_null_char
   call c_f_strpointer( c_loc(s), ps, nchars=22_c_size_t )
   print *, "ps: ", ps
   print *, "len_trim(ps) = ", len_trim(ps), "; expected is 22"
end

[Steve_Lionel]

@FortranFan 's example compiles successfully with ifx.

[Ron_Green]

Thanks for sending this. The bug ID is CMPLRLLVM-43336

[Robert_van_Amerongen]

Thank all of you for your help. As FortranFan suggests, the routine was written to have it available as soon as possible. The new standard has a series of new procedures that are very, very, welcome. The C-string helpers are one of these.

In the mean time I found some other ICE's . I will take a cl.oser look to see if these are related with the ICE here; I will let you know.

Robert

[Ron_Green]

this bug is fixed in 2023.2.0

[Robert_van_Amerongen]

Good to read.

Thanks for letting me know.