Re: gtk3 vs. current: nearbyint

[Thomas Klausner <wiz%NetBSD.org@localhost>]

On Sat, Feb 09, 2013 at 05:44:18PM +0000, Patrick Welche wrote:
> On Sat, Feb 09, 2013 at 05:42:55PM +0100, Thomas Klausner wrote:
> > On Sat, Feb 09, 2013 at 04:38:25PM +0000, Patrick Welche wrote:
> > > On Sat, Feb 09, 2013 at 03:38:07PM +0100, Thomas Klausner wrote:
> > > > gtk3 doesn't build on -current.
> 
> > Yes, this only started occurring after my updated to a Feb 7 snapshot;
> > my one before that was 20130116.
> 
> > To generate a diff of this commit:
> > cvs rdiff -u -r1.58 -r1.59 src/include/math.h
> 
> It'll take me a while to update - it looks as though before that commit,
> we simply didn't have nearbyint(), so gtk's configure didn't find it, and
> the gtk provided version of nearbyint() in fallback-c89.c was used.
> 
> The commit to include/math.h includes
> 
> +double nearbyint(double);
> 
> So, do we really have nearbyint(), or is it just in the header?

It looks like we only have the header:
> cat test.c
#include <math.h>
#include <stdio.h>

int main() {
        printf("%f", nearbyint(3.5));
}
> gcc test.c
/var/tmp//ccWfMjjI.o: In function `main':
ii.c:(.text+0xd): undefined reference to `nearbyint'
> gcc test.c -lm
/var/tmp//ccmUqM0p.o: In function `main':
ii.c:(.text+0xd): undefined reference to `nearbyint'

Should the math.h change be backed out?

Also, for whatever reason the prototype seems to conflict with the one
the configure script expects (see my first mail).
 Thomas