[LCP]Rounding questions

[Steve Baker]

Chuck Martin <nrocinu at myrealbox.com> wrote:
> On Thu, Sep 05, 2002 at 12:36:39PM -0500, Steve Baker wrote:
> >   It seems perplexing, but the problem is that round is not prototyped in
> > math.h, or anywhere for that matter.  This is clearly a glibc maintainer
> > issue.  Add to your program at the top:
> > 
> > double round(double);
>
> Thank you!  This works.  Do you happen to know why rint() works without
> doing this?  I can't seem to find a prototype for it anywhere, including
> math.h.

  Augh... I _always_ forget about features.h.  round() is defined after all,
but only if you have _ISOC99_SOURCE defined at the very very top of your C
file before any #includes (or _GNU_SOURCE, which turns on other things as
well).  If you look in /usr/include/features.h you'll find all the "features"
you can select from.

  rint() is a BSD'ism thus it is enabled by default.  round() is a C99 feature
thus you must enable its use.

  I was using cpp to see if round was being defined, but didn't see it thus I
assumed it wasn't defined anywhere, but was in fact, only #defined out of
existence.  I overlooked it while rgrep'ing for it because all math.h function
calls are prototyped like:

__MATHCALL (round,, (_Mdouble_ __x));

  They don't make it easy to track down, but there you go.  Stick:

#define _ISOC99_SOURCE

  At the very top of your C file and you're C99 ready.

> Now this brings me to the other question I asked when I started this
> thread.  How portable is this?  The man page says it conforms to C99,
> but I'm not up on all of the standards and where they're all used.
> Will using this instead of my own rounding routine cause problems in
> portability, and if so, to what extent?

  Any system that conforms to C99 should be OK, and most should, so I imagine
it's portable.  You may have to jump through a hoop or two like this to get it
to work though.

								- Steve