Re: [Tracker] GLib-CRITIAL warnings about g_timer_continue

[Martyn Russell <martyn lanedo com>]

On 23/11/13 16:07, Carlos Garnacho wrote:
> Hey Philip,
>
> On sáb, 2013-11-23 at 10:46 +0100, Philip Van Hoof wrote:
> > Hi guys,
> >
> > Logs from a systemd booted system on Tracker 0.14:
> >
> > Nov 22 16:41:47 localhost dbus-daemon[901]: (tracker-miner-fs:948):
> > GLib-CRITICAL **: g_timer_continue: assertion `timer->active == FALSE'
> > failed
> > Nov 22 16:41:47 localhost tracker-miner-fs[948]: GLIB CRITICAL ** GLib -
> > g_timer_continue: assertion `timer->active == FALSE' failed
> >
> > Occurences in tracker-miner-fs's code:
> >
> > libtracker-miner/tracker-crawler.c:
> >         g_timer_continue (crawler->priv->timer);
> >
> > libtracker-miner/tracker-miner-fs.c:
> >           } else if (fs->priv->extraction_timer_stopped) {
> >                   g_timer_continue (fs->priv->extraction_timer);
> >                   fs->priv->extraction_timer_stopped = FALSE;
> >           }
> >
> > Can we in the situation that happened simply activate the timer instead
> > of using continue?
>
>
> TrackerMinerFS initialization indeed seemed a bit broken, the timers
> were created (so marked as already running), but those boolean guards
> were set to TRUE, so there was a disagreement at first.
>
> I pushed a small fix for that on master an 0.16. Still, the idle logic
> is somewhat confusing, which doesn't help much in timing it, I hope I
> didn't miss anything :)

I always thought this was a GLib bug actually, but I could be wrong. 
Been a while since I looked at this code. It's been around for a while 
this issue. We did work around it at some point I thought.

Thanks for handling Carlos,

--
Regards,
Martyn

Founder & Director @ Lanedo GmbH.
http://www.linkedin.com/in/martynrussell