Re: [fortran, patch] Enable Hollerith constant and character array in format tag.

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

On Tue, Mar 29, 2005 at 01:34:40AM +0200, Steven Bosscher wrote:
> On Monday 28 March 2005 16:17, Feng Wang wrote:
> > 	* arith.c (gfc_hollerith2int, gfc_hollerith2real,
> > 	gfc_hollerith2complex, gfc_hollerith2character): New functions.
> 
> All of these look wrong to me.  If I understand Holleriths correctly,
> their value is a bitwise copy of the string to the target format of
> whatever they're supposed to represent.  But you are copying the string
> value of the Holleriths into an mpz_t, a single mpfr_t, or a pair of
> mpfr_t.  That can't be right.
> 
> Gr.
> Steven

I don't think he's storing the Hollerith in the integer's mpz_t,   
the real's  single mpfr_t, or complex's 2 mpfr_t.  A closer
look shows that he is storing the KIND number of bytes in the
result's string field.

gfc_expr *
gfc_hollerith2int (gfc_expr * src, int kind)
{
   gfc_expr *result;
   int len;

   len = src->value.character.length;

   result = gfc_get_expr ();

The next few lines set the type and kind for integer, but
also sets the from_H bit to true.

   result->expr_type = EXPR_CONSTANT;
   result->ts.type = BT_INTEGER;       
   result->ts.kind = kind;
   result->where = src->where;
   result->from_H = 1;
  
This if () now determines the total number of bytes to use

   if (len > kind)
     {
       gfc_warning ();
       len = kind;
     }

Here he is saving the KIND number of bytes into result's
value.character.string field.  In essences, he isn't doing
a conversion.

   result->value.character.string = gfc_getmem (len + 1);
   memcpy (result->value.character.string, src->value.character.string, len);
   result->value.character.string[len] = '\0';
   result->value.character.length = len; 

   return result;
 }

-- 
Steve