Re: [Tracker] r978 unable to start up when invoked indirectly

[jamie <jamiemcc blueyonder co uk>]

On Tue, 2007-11-06 at 15:09 +0200, Tshepang Lekhonkhobe wrote:
> On Nov 6, 2007 2:27 PM, jamie <jamiemcc blueyonder co uk> wrote:
> > On Tue, 2007-11-06 at 12:55 +0200, Tshepang Lekhonkhobe wrote:
> > > On Nov 6, 2007 12:43 PM, jamie <jamiemcc blueyonder co uk> wrote:
> > > > On Tue, 2007-11-06 at 12:39 +0200, Tshepang Lekhonkhobe wrote:
> > > > > On Oct 31, 2007 3:56 PM, Tshepang Lekhonkhobe <tshepang gmail com> wrote:
> > > > > > On 10/31/07, jamie <jamiemcc blueyonder co uk> wrote:
> > > > > > > On Wed, 2007-10-31 at 15:30 +0200, Tshepang Lekhonkhobe wrote:
> > > > > > > > On 10/31/07, jamie <jamiemcc blueyonder co uk> wrote:
> > > > > > > > > On Wed, 2007-10-31 at 15:00 +0200, Tshepang Lekhonkhobe wrote:
> > > > > > > > > > Hi,
> > > > > > > > > > r978 has completed indexing my data and works fine until I restart it
> > > > > > > > > > indirectly (IE via tracker-stats and deskbar-applet). When I start it
> > > > > > > > > > directly (trackerd) it's no trouble. Here's where it crashes (with no
> > > > > > > > > > further messages):
> > > > > > > > > >
> > > > > > > > > > [snip]
> > > > > > > > > > 30 Oct 2007, 16:53:34:132 - Using Sqlite version 3.4.2
> > > > > > > > > > 30 Oct 2007, 16:53:34:132 - Loading prepared queries...
> > > > > > > > > > 30 Oct 2007, 16:53:34:133 - File loaded in  0.189000 ms
> > > > > > > > > > 30 Oct 2007, 16:53:34:135 - loaded sql file sqlite-cache.sql
> > > > > > > > > > 30 Oct 2007, 16:53:34:136 - Database closed for thread main
> > > > > > > > > > [end]
> > > > > > > > > >
> > > > > > > > > > Note that this problem did not exist before this revision, except some
> > > > > > > > > > months ago, but then somehow got fixed (and now unfixed :-)
> > > > > > > > >
> > > > > > > > > the integrity check at startup is likely causing a timeout for the
> > > > > > > > > clients
> > > > > > > > >
> > > > > > > > > trackerd should not crash though...
> > > > > > > >
> > > > > > > > It crashes (disappears from lsof's reach) after a few minutes, without
> > > > > > > > warning :-( and this happened on two separate machines.
> > > > > > >
> > > > > > > ok I will investigate it - you sure it does not crash when started
> > > > > > > beforehand?
> > > > > >
> > > > > > Tried a few times actually and the complaint still stands. Changing
> > > > > > tracker.cfg to default don't help.
> > > > >
> > > > > Thanks. Problem is fixed a few commits later...
> > > > well maybe - you need to kill -9 trackerd and then restart to test as we
> > > > only do corruption check if trackerd wes not shutdown cleanly
> > >
> > > Something else definitely changed because I've done that already (kill
> > > -9) with trackerd not dying on me when externally invoked.
> >
> > why did you do kill -9?
> >
> > its very important we makes sure trackerd always exits cleanly (within
> > 10 secs)
> >
> > If you know anywhere where it hangs or does not exit cleanly can you let
> > me know?
> >
> > best to run it in gdb so you can get me a backtrace where it hangs
>
> There must have been a misunderstanding because I thought you wanted
> me to check if it will hang if restarted after being murdered (kill
> -9). It runs fine here (no crashing).

yeah thats fine

just pointing out that it should not be necessary to do kill -9 ever


if it is let me know

>