On Jan 19 at 15:46, Thomas Dickey spoke:
> On Mon, 19 Jan 2004, Hanspeter Roth wrote:
>
> > So should the invokation of getloadavg() be moved to a separate C
> > module and then call the C function that calls getloadavg() from a
> > C++ module?
>
> That might be a better design - it would at least get rid of the problem
> of C++ breaking the configure test. Since it's not standard, there's
> no guarantee that its prototype is the same across all implementations.
>
> The various Linux and *BSD platforms tend to make random uncoordinated
> changes to the legacy BSD functions, sprinking const's in their
> prototypes.
It's for icewm. The Linux part is covered by another #ifdef anyway.
So I'm also considering some AC_TRY_LINK check:
AC_MSG_CHECKING([for getloadavg])
AC_TRY_LINK([#include <stdlib.h>],
[double loadavg[3]; getloadavg(loadavg, 3) == 0;],
[ AC_MSG_RESULT(yes)
AC_DEFINE(HAVE_GETLOADAVG, 1, [getloadavg() is available])],
[ AC_MSG_RESULT(no) ])
I guess this should avoid conflicts with uncompatible getloadavg()s.
(My primary intend is to support Free-, Net- and OpenBSD.)
There's also AC_FUNC_GETLOADAVG. This would define C_GETLOADAVG if
run by g++ 3.3. But I don't know how it behaves on platforms that
have getloadavg() but on which I have no access to...
-Hanspeter
