From: mschulter AT  DOT value DOT net (M. Schulter)
Newsgroups: comp.os.msdos.djgpp
Subject: pb with long double
Date: 15 Aug 1997 00:49:32 GMT
Organization: Value Net Internetwork Services Inc.
Lines: 72
Message-ID: <5t092s$hsi$1@vnetnews.value.net>
NNTP-Posting-Host: value.net
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Precedence: bulk

> I am a french student and I have little problem with GCC
> Here is my program :

> #include <math.h>
> main()
> {
>  long double a;
>  a=PI;
>  printf("%22.20Lf",a);
> }

> The result is 3.14159265358979311600
>                               wrong

> There is only 15 good digits instead of 19
> ( PI is defined with more than 20 digits im math.h )

Hi, there. Here's a program which _may_ illustrate the problem and a
possible solution by returning the desired value with long double
precision:

#include <stdio.h>
#include <math.h>
#define L_PI 3.14159265358979323846L

int main(void)
{
 long double a;
 a=L_PI;
 printf("%22.20Lf",a);
 return(0);
}

At least with the outdated GCC 2.7.2 (DJGPP 2.00) that I'm regrettably
still using <grin>. it seems that a long double quantity must be
followed by the designator 'L' to be treated as a long double. When I
define my own L_PI as 3.14159...846L, running the revised program in
Emacs gives the result:

3.14159265358979323851

This result is correct to the first 19 digits, about what I would
expect from a long double.

Here's another example, this time focusing on magnitude:

#include <stdio.h>
#include <math.h>

int main(void)
{
 long double a;
 a=3.11e4000L;
 printf("%22.20Le",a);
 return(0);
}

Note that this works to around 19-digit precision, but if I rewrite
the assignment to:

a=3.11e4000;

without that 'L', then the output from GCC 2.7.2 is 'Inf' -- since the
limits of a double have been exceeded.

Most respectfully,

Margo Schulter
mschulter AT value DOT net