[Sugar-devel] [ASLO] Release Clock-17

[James Cameron]

The problem followed the SD card.

Booting from NAND Flash on two different systems, the
GStreamer-CRITICAL warnings and related silence occur roughly one out
of every 30 starts from journal.

So there is a race condition of some sort, but it affects some laptops
more than others, because of storage I/O timing.

It might have been there already, so unless you can reproduce it,
don't worry about it.

On Thu, May 14, 2015 at 12:52:37PM +1000, James Cameron wrote:
> Thanks.  Applied both.  Continue to show segmentation faults, the
> GStreamer-CRITICAL warnings, a lock up CPU loop (not fixed by stop in
> frame), and rarely:
> 
> *** glibc detected *** python2: malloc(): memory corruption: 0x0a2dcb68 ***
> Terminated by signal 11, pid 2688 data (None, <open file '<fdopen>', mode 'w' at 0x993a9c0>, '3d0b3c7e93a6309a37688bcf60186a2ed9533795')
> 
> The same hardware is not showing any other signs of distress.
> 
> But it is running from SD card, so I'll try from internal NAND Flash.
> 
> On Wed, May 13, 2015 at 12:40:11PM -0300, Gonzalo Odiard wrote:
> > I have pushed two changes to the clock activity repo
> > and send a pr to sugar-toolkit-gtk3 [1]
> > 
> > The pr avoid checking if the espeak plugin is installed every time
> > the enabled() function is called. I suppose that could remove the scanning.
> > 
> > I couldn't reproduce the crash here, testing in sugar-build on F21,
> > in a xo-1 with 13.2.4 and in a vm with ubuntu.
> > 
> > Gonzalo
> > 
> > [1] [1]https://github.com/sugarlabs/sugar-toolkit-gtk3/pull/215
> > 
> > On Tue, May 12, 2015 at 7:21 PM, James Cameron <[2]quozl at laptop.org> wrote:
> > 
> >     On Tue, May 12, 2015 at 06:22:42PM -0300, Gonzalo Odiard wrote:
> >     > I think I found the problem. 
> >     > Could you confirm replacing the attached file in the activity dir?
> > 
> >     Thanks, that's much better.  The warning about rate and the delay
> >     before speaking have been fixed.
> > 
> >     Rarely, about one in 10 times, it doesn't speak, and reports in log
> > 
> >     (sugar-activity:2492): GStreamer-CRITICAL **: gst_memory_new_wrapped:
> >     assertion `offset + size <= maxsize' failed
> > 
> >     (sugar-activity:2492): GStreamer-CRITICAL **: gst_buffer_insert_memory:
> >     assertion `mem != NULL' failed
> > 
> >     Rarely, about one in 5 times, it does segmentation fault.  I've
> >     captured two core dumps by adding "ulimit -c unlimited" to
> >     ~/.sugar/debug.  Both showed similar stack traces; a thread busy doing
> >     an fts_open and fts_sort at the same time as the main thread doing
> >     something else.
> > 
> >     [3]http://dev.laptop.org/~quozl/z/1YsIKF.txt
> > 
> >     Speculation: the Gst.init() in src/sugar3/speech.py creates a thread
> >     which is racing with something in the main thread.  fts_open and
> >     fts_sort are necessary calls for scanning plugins.
> > 
> >     Sometimes the race condition corrupts pointers and we get the silence
> >     with GStreamer-CRITICAL messages.  Sometimes the corruption causes a
> >     segmentation fault.
> > 
> >     Perhaps _GstSpeechPlayer should wait for good Gst.State before the
> >     next call into Gst?
> > 
> >     I remember recent work on GStreamer plugin scan initialisation in
> >     Sugar, but I can't find it yet.
> > 
> >     --
> >     James Cameron
> >     [4]http://quozl.linux.org.au/
> > 
> > --
> > Gonzalo Odiard
> > 
> > SugarLabs - Software for children learning 
> > 
> > References:
> > 
> > [1] https://github.com/sugarlabs/sugar-toolkit-gtk3/pull/215
> > [2] mailto:quozl at laptop.org
> > [3] http://dev.laptop.org/~quozl/z/1YsIKF.txt
> > [4] http://quozl.linux.org.au/
> 
> > _______________________________________________
> > Sugar-devel mailing list
> > Sugar-devel at lists.sugarlabs.org
> > http://lists.sugarlabs.org/listinfo/sugar-devel
> 
> 
> -- 
> James Cameron
> http://quozl.linux.org.au/

-- 
James Cameron
http://quozl.linux.org.au/