svn commit: r209110 - in head/lib/msun: . src

[David Schultz]

On Fri, Dec 03, 2010, Kostik Belousov wrote:
> On Thu, Dec 02, 2010 at 02:00:10PM -0500, David Schultz wrote:
> > On Thu, Dec 02, 2010, Kostik Belousov wrote:
> > > For __isnanf, libc exports __isnanf at FBSD_1.0, and libm exports
> > > __isnanf at FBSD_1.2. I suspect that we could export both
> > > __isnanf at FBSD_1.0 and __isnanf at FBSD_1.2 from libc.
> > 
> > I like the idea of adding an __isnanf at FBSD_1.2 alias to libc for
> > the benefit of people running -CURRENT.
> 
> Ok, draft of the change is below, I am not sure is it worth the churn.
> 
> I was unable to make the ld support two versions of the same symbol.
> lib/msun would benefit from elimination of isnan sources if this is
> committed.
[...]
> --- /dev/null
> +++ b/lib/libc/gen/isnan_compat.c
> @@ -0,0 +1,17 @@
> +/*
> + * The file is in public domain.
> + * Written by Konstantin Belousov <kib at FreeBSD.org>
> + *
> + * $FreeBSD$
> + */
> +
> +#include <math.h>
> +
> +__sym_compat(__isnanf, __isnanf_compat, FBSD_1.0);
> +
> +int
> +__isnanf_compat(float f)
> +{
> +
> +	return (__isnanf(f));
> +}

Thanks.  Hmm... if it ultimately entails replacing a small amount
of cruft in libm with a small amount of cruft in libc, it's
probably not worth the churn.