On Mon, Sep 03, 2007 at 08:03:26PM -0400, Mike Frysinger wrote:
> +]], [[int test = __NR_$1;]])],v
> + [sysnum="__NR_$1"],[sysnum=""])
> + if test "x$sysnum" != "x" ; then
> + AC_MSG_RESULT([__NR_$1])
> + else
> + case $host_cpu in
> + $2
> + *) sysnum=-1;;
> + esac
> + if test "x$sysnum" = "x-1" ; then
> + AC_MSG_RESULT([unable to detect])
> + else
> + AC_MSG_RESULT([$sysnum])
> + fi
> + fi
> + AC_DEFINE_UNQUOTED([SYS_$1], $sysnum, [Fallback syscall number for $1 in place of __NR_$1])
> + ])
> +])
It means:
#define SYS_<foo> -1
is case the syscall <foo> is unsupported by system during build time.
That's bad. It doesn't make sense to build utils that are completely
useless on arbitrary system.
There should be two choices:
1) fallback logic (temporary for "newish" syscalls)
- we build independently on the current kernel/glibc
- if SYS_<foo> is undefined by system, but we have fallback:
AC_MSG_WARN([your system does not support SYS_$1, using $sysnum])
2) AC_CONDITIONAL(HAVE_SYSCALL_$1 ........)
I think we still need things like:
if HAVE_SCHED_GETAFFINITY
usrbinexec_PROGRAMS += taskset
man_MANS += taskset.1
endif
Your old patch "check for existence of sched_getaffinity" still
makes a lot of sense for archs where this syscall is not implemented.
Karel
--
Karel Zak <kzak@xxxxxxxxxx>
-
To unsubscribe from this list: send the line "unsubscribe util-linux-ng" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
