Re: [Linphone-developers] [PATCH] Fix busy cycle if date change

[damico]

Hi Aymeric,
unfortunately on my uClibc version (9.27) CLOCK_MONOTONIC was not 
implemented yet. It was introduced from 9.28. So, now I must use my 
patch and I can't test your (best) solution; but I'm quite sure that 
CLOCK_MONOTONIC is the real fix and my code just a work around.

A note: maybe should be better check if CLOCK_MONOTONIC is implemented 
before use it as exclusive source of time in linux system...

Regards

--Michele
damico ha scritto:
> Hi Aymeric,
>
> I'm using mediastreamer 2.2.0 and, as you said, it uses CLOCK_REALTIME 
> instead of CLOCK_MONOTONIC.
>
> I'll try to change it and check if the problem disappears... I can't 
> upgrade mediastreamer now but I'll upgrade it to the last version at 
> soon as possible.
>
> What is the last version of mediastreamer? There is a "bugs list" 
> and/or "new features list" for the last version of mediastreamer?
>
> Regards
>
> --Michele
>
> Aymeric Moizard ha scritto:
> > On Wed, 6 May 2009, damico wrote:
> >
> > > Hi all,
> > > i found a issue when date change and a ticker running:
> > >
> > > If the date move backward the ticker thread become a busy cycle. On 
> > > my embedded solution that use an ntp server that issue causes a lot 
> > > of disasters :)
> >
> > This was fixed by using CLOCK_MONOTONIC instead of CLICK_REALTIME
> > some month ago I think. Did you experienced this with the latest
> > version? Can you check get_cur_time in msticker.c?
> >
> > tks,
> > Aymeric MOIZARD / ANTISIP
> > amsip - http://www.antisip.com
> > osip2 - http://www.osip.org
> > eXosip2 - http://savannah.nongnu.org/projects/exosip/
> >
> > > I'm proposing that code to solve the problem... any comments?
> > >
> > > Regards
> > >
> > > --Michele
> > >
> > > Index: mediastreamer2/src/msticker.c
> > > ===================================================================
> > > --- mediastreamer2/src/msticker.c    (revision 463)
> > > +++ mediastreamer2/src/msticker.c    (working copy)
> > > @@ -331,6 +331,8 @@
> > >
> > > #ifndef WIN32_TIMERS
> > >
> > > +#define _MAX_DELAY_FOR_SYNC_FORCE_MS (1*60*1000) /* 1 min */
> > > +
> > > void * ms_ticker_run(void *arg)
> > > {
> > >    uint64_t realtime;
> > > @@ -352,9 +354,19 @@
> > >        run_graphs(s,s->execution_list,FALSE);
> > >        s->time+=s->interval;
> > >        while(1){
> > > -            
> > > realtime=s->get_cur_time_ptr(s->get_cur_time_data)-s->orig;
> > > +            uint64_t now = s->get_cur_time_ptr(s->get_cur_time_data);
> > > +            if (now < s->orig){
> > > +                ms_warning("Back from the future....Sync");
> > > +                s->orig = now - s->time;
> > > +            }
> > > +            realtime = now - s->orig;
> > > +            diff = s->time - realtime;
> > > +            if (diff < -_MAX_DELAY_FOR_SYNC_FORCE_MS || diff > 
> > > _MAX_DELAY_FOR_SYNC_FORCE_MS){
> > > +                ms_warning("... Too much diff (does date be changed 
> > > forward?)....Sync");
> > > +                s->orig=now - s->time;
> > > +                diff = 0;
> > > +            }
> > >            ms_mutex_unlock(&s->lock);
> > > -            diff=s->time-realtime;
> > >            if (diff>0){
> > >                /* sleep until next tick */
> > >                sleepMs((int)diff);
> > >
> > >
> > >
> > > _______________________________________________
> > > Linphone-developers mailing list
> > > address@hidden
> > > http://lists.nongnu.org/mailman/listinfo/linphone-developers
>
>
>
> _______________________________________________
> Linphone-developers mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/linphone-developers