[Olsr-users] 0.5.6 Routes disappear after 4 mins of uptime, then all OK - suspect clock sync
Wed Sep 3 15:52:55 CEST 2008
mmh. CLOCK_MONOTONIC is not supported by uclibc AFAIK.
"Bernd Petrovitsch" <(spam-protected)> schrieb im Newsbeitrag
> On Die, 2008-09-02 at 21:16 +0200, John Hay wrote:
>> On Tue, Sep 02, 2008 at 07:24:01PM +0200, Bernd Petrovitsch wrote:
>> > Not really:
>> > gettimeofday(2) and time(2) return the current date+time (read: that
>> > what the device thinks it is) or some representation of it.
>> > The problem with that is that it is not continuous because
>> > ntp/ntpdate/rdate/the admin/daylight-saving-time/... may change speed
>> > and value at any time (and I purposely ignore leap seconds).
>> > These values are great for syslog and other time values which refer to
>> > the rest of the planet.
>> > For timeouts in select/poll/sleep/timers in general one wants a steady
>> > clock without jumps. And times(2) is the only portable sys-call I know
>> > of.
>> What about clock_gettime(2)? According to the FreeBSD man page it is
> I assume you mean the CLOCK_MONOTONIC "version".
> Why not?
>> part of IEEE Std 1003.1b-1993 (``POSIX.1'').
> Sorry;-), but times(2) is also in SuSv3:
> Sven-Ola (or anyone else;-), do we want/have librt.(a|so) on openwrt and
> other smallish hardware/systems?
> At least on my Fedora-9, I need "-lrt" to be able to link it.
> Or did I overlook something?
>> > BTW do we have requirements on precision?
>> clock_gettime(2) give down to nanosecond level, which should be ok for
>> a while I guess. :-)
> IMHO for sure. But it is needed?
> Firmix Software GmbH http://www.firmix.at/
> mobil: +43 664 4416156 fax: +43 1 7890849-55
> Embedded Linux Development and Services
> Olsr-users mailing list
More information about the Olsr-users