Re: patch for wanna-build (sysinfo/getloadavg)

[Robert Millan <zeratul2@wanadoo.es>]

Thanks everyone for the tips. I'm sending the fix to Roman Hodek.

On Fri, Aug 08, 2003 at 08:55:51PM +0100, Roger Leigh wrote:
> Robert Millan <zeratul2@wanadoo.es> writes:
> 
> > On Thu, Jul 24, 2003 at 12:14:57AM +0100, Roger Leigh wrote:
> >> Robert Millan <zeratul2@wanadoo.es> writes:
> >> 
> >> > Some days ago I posted this small patch for building wanna-build on
> >> > GNU, and asked for revision, but didn't get any response.
> >> >
> >> > It attempts to replace the linux-specific sysinfo() with getloadavg(),
> >> > but there's a type conversion issue and i'd like to make sure i did it
> >> > correctly, please could anyone verify this:
> >> >
> >> > --- wanna-build.old/buildd-mail-wrapper.c	1999-07-21 14:53:22.000000000 +0200
> >> > +++ wanna-build/buildd-mail-wrapper.c	2003-07-15 21:06:59.000000000 +0200
> >> [...]
> >> 
> >> This seems rather complex!  Why not just do something like:
> >>                 waittime = (long) (loadavg[0] / 16384) * 6 + 20;
> >> 
> >> If you don't need SI_LOAD_SHIFT (Linux-specific??), it can probably be
> >> even simpler.
> >
> > Can I cast a double to long like that? What kind of truncating happens if
> > loadavg[0] is bigger than LONG_MAX? Does a division with 2^14 garantee it
> > will fit nicely?
> 
> I've written a simple test program.  It seems any cast from
> double->long will set LONG_MIN/LONG_MAX if it would result in an
> double->overflow.  Thus,
> 
> +		if (loadavg[0] > LONG_MAX)
> +			load = LONG_MAX;
> +		else
> +			load = (long) loadavg[0];
> 
> can be replaced with
> 
> load = (long) loadavg[0].
> 
> This was tested on Debian unstable (Linux 2.4.21 with glibc 2.3.1).
> For some reason, I only ever got LONG_MIN to be returned, even when
> the overflow was positive!  I'm not sure if this is a problem with my
> code, or a glibc bug, or whatever.
> 
> Here's the test program.  You'll need to link with -lreadline.
> 
> #include <stdio.h>
> #include <stdlib.h>
> #include <limits.h>
> #include <errno.h>
> 
> #include <readline/readline.h>
> 
> int main(void)
> {
>   char *input = NULL;
>   double big;
>   long fixed;
> 
>   while(1)
>     {
>       if (input)
> 	{
> 	  free(input);
> 	  input = NULL;
> 	}
>       input = readline("Enter a number: ");
> 
>       big = strtod(input, NULL);
> 
>       if (errno == ERANGE)
> 	{
> 	  fprintf(stderr, "Number is out of range: ");
> 	  if (big == 0)
> 	    fprintf(stderr, "FP underflow\n");
> 	  else
> 	    fprintf(stderr, "FP overflow\n");
> 	  continue;
> 	}
> 
>       printf("double: %g\n", big);
> 
>       fixed = (long) big;
> 
>       if (fixed == LONG_MIN || fixed == LONG_MAX)
> 	{
> 	  fprintf(stderr, "Number is out of range for type \"long\": ");
> 	  if (fixed == LONG_MIN)
> 	    fprintf(stderr, "LONG_MIN");
> 	  else
> 	    fprintf(stderr, "LONG_MAX");
> 	  fprintf(stderr, ": %ld\n", fixed);
> 	  continue;
> 	}
> 
>       printf("long: %ld\n", fixed);
>     }
> 
>   return 0;
> }
> 
> 
> -- 
> Roger Leigh
> 
>                 Printing on GNU/Linux?  http://gimp-print.sourceforge.net/
>                 GPG Public Key: 0x25BFB848 available on public keyservers
> 
> 
> -- 
> To UNSUBSCRIBE, email to debian-hurd-request@lists.debian.org
> with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
> 

-- 
Robert Millan

"[..] but the delight and pride of Aule is in the deed of making, and in the
thing made, and neither in possession nor in his own mastery; wherefore he
gives and hoards not, and is free from care, passing ever on to some new work."

 -- J.R.R.T, Ainulindale (Silmarillion)