Bug 2442

Additional Comment #1 From John Ellson (ellson@research.att.com)  on 2003-06-13 

sincos() is not very portable.   Here is the configure test and a sincos
macro that we've used with success in graphviz:

--------------------- configure test -----------------------

AC_MSG_CHECKING(if have working sincos())
AC_TRY_RUN([
        #include <stdio.h>
        #include <assert.h>
        #define PI 3.14159265358979323846
        extern void sincos(double x, double *s, double *c);
        int main () {
                double sinx=0.0, cosx=1.0;
                sincos(PI/2.0,&sinx,&cosx);
                assert(sinx>0.9999999);
                assert(sinx<1.0000001);
                assert(cosx<0.0000001);
                assert(cosx>(-0.0000001));
                return 0;
        }]
        ,
        AC_MSG_RESULT(yes)
        AC_DEFINE_UNQUOTED(HAVE_SINCOS,1,[Define if libm provides a *working*
sincos function])
        ,
        AC_MSG_RESULT(no)
        ,
        AC_MSG_RESULT(no - assumed because cross-compiling)
        )


------------------ macro ---------------------

#ifndef HAVE_SINCOS
void sincos(x,s,c) double x,*s,*c; { *s = sin(x); *c = cos(x); }
#else
extern void sincos(double x, double *s, double *c);
#endif

Someone would have to translate the autotool test above to an Imake test
of some sort, if there isn't one already present.  Shouldn't be that
difficult.

We were tracking this in our bugzilla, but it was low priority enough
that we might never get around to doing it, but I figured someone else
in the community might want to take a crack at it, so it'd be useful
to log it here perhaps.

Original bug report:  https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=90053

glxgears properly lives in Mesa, so i wouldn't bother with either the configure
test or the Imake test since Mesa uses neither.

I suggest closing this bug.  There no /real/ benefit in using sincos().  Why bother?

last thing we want is people thinking gears is a benchmark.