Re: iostat

[Steve Kargl <sgk at troutmask dot apl dot washington dot edu>]

On Wed, Aug 23, 2006 at 09:42:50AM +0300, Erik Edelmann wrote:
> 
> What about assigning the value of errno to iostat, and then if the user
> wants to know what an error code means on his system, he can do
> something like
> 
> 
> #include <stdio.h>
> int main () { puts(strerror(<the error code>)); return 0; }
> 
> 
> It requires some basic C knowledge of course, so it's not an ideal
> solution, but if someone is desperate enough, it's still doable.
> 

If we do this (and I'm not necessarily opposed to better error
messages), then I think we should add new (nonstandard) intrinsic
procedure, say, iostat_to_msg().  This would be much easier to
document, and avoids having to list errno values in gfortran.texi,
and does require gfortran user to gravel through C code/headers.

program test
  integer ios
  character(len=80) msg
  open(unit=1,file='yada',iostat=ios)
  if (ios /= 0) then
     call iostat_to_msg(ios, msg)
  end if
end program test.

Where we have 

  interface
    subroutine iostat_to_msg(ios, msg)
      integer, intent(in) :: ios
      character(len=*), intent(out) :: msg
    end subroutine iostat_to_msg
  end interface

and

void
iostat_to_msg(GFC_INTEGER_4 *iostat, char *msg, gfc_charlen_type len)
{
   strncpy(msg, strerror(*iostat), len-1);
}

-- 
Steve