On Tue, 09 Mar 2021 19:09:01 +0100
Yann Droneaud <ydroneaud@opteya.com> wrote:

> Hi,
> 
> Le mardi 09 mars 2021 à 12:47 +0100, Lukasz Majewski a écrit :
> 
> diff --git a/sysdeps/unix/sysv/linux/tst-ntp_gettime.c
> b/sysdeps/unix/sysv/linux/tst-ntp_gettime.c
> new file mode 100644
> index 0000000000..5a1261fd35
> --- /dev/null
> +++ b/sysdeps/unix/sysv/linux/tst-ntp_gettime.c
> @@ -0,0 +1,49 @@
> +/* Test for ntp_gettime
> +   Copyright (C) 2021 Free Software Foundation, Inc.
> +   This file is part of the GNU C Library.
> +
> +   The GNU C Library is free software; you can redistribute it and/or
> +   modify it under the terms of the GNU Lesser General Public
> +   License as published by the Free Software Foundation; either
> +   version 2.1 of the License, or (at your option) any later version.
> +
> +   The GNU C Library is distributed in the hope that it will be
> useful,
> +   but WITHOUT ANY WARRANTY; without even the implied warranty of
> +   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
> +   Lesser General Public License for more details.
> +
> +   You should have received a copy of the GNU Lesser General Public
> +   License along with the GNU C Library; if not, see
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#include <time.h>
> +#include <stdlib.h>
> +#include <sys/timex.h>
> +#include <support/check.h>
> +#include <support/xtime.h>
> +
> +#ifndef NTP_GETTIME_SYSCALL
> +# define NTP_GETTIME_SYSCALL ntp_gettime
> +#endif
> +
> +# define STR(__s) #__s
> +
> +static int
> +do_test (void)
> +{
> +  struct timespec tv_now;
> +  struct ntptimeval ntv;
> +
> +  int ret = NTP_GETTIME_SYSCALL (&ntv);
> +  if (ret == -1)
> +    FAIL_EXIT1 (STR(NTP_GETTIME_SYSCALL)" failed: %m\n");
> +
> +  tv_now = xclock_now (CLOCK_REALTIME);
> +
> +  if (tv_now.tv_sec != ntv.time.tv_sec)
> +    FAIL_EXIT1 ("ntp_gettime provided wrong time!\n");
> +
> 
> This can fail if ntv.time.tv_usec is somewhere around 999999µs and the
> process is scheduled before completing xclock_now().
> 
> It would be better to try again a few times.

If we want to avoid such error - maybe it would be better to convert
struct ntptimeval's struct timeval time member to struct timespec and
then use timespec_sub() to get the time difference.

Then this difference could be assessed in
support_timespec_check_in_range() with some allowed percentage
variation (like 10%).


> 
> Regards.
> 




Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma@denx.de