[Sugar-devel] 11.3.0 release candidate 4 (build 883) released

[James Cameron]

On Mon, Oct 31, 2011 at 06:41:21PM +0000, Gary Martin wrote:
> I've just been testing Telescope-12 on the XO-1.75 running the new 883 build, first the good news, the camera is showing up correctly in Record and Telescope, fab! :) Now the not so hot:
> 
> - While in Telescope-12, clicking the digital zoom toolbar button completely locked up the XO-1.75 (re-testing with an XO-1 the digital zoom works ok).
> 
> - After some waiting and many attempted keyboard shortcuts, I decided to power down. Clicking the power hardware button did not trigger the usual shut down screen during this particular lockup case; clicking the power button twice (assuming it might just be the gfx that had locked-up) had no effect; finally I held down the power button for some seconds until the power led flashes and then the hard powers off.

Speculation.  A kernel panic had happened.  This is recorded nowhere
except the serial log.  The effect is to prevent any further activity.

> - On reboot the Journal now shows 3 corrupt entries (see attached screen shot) with no name and unknown metadata. None of the three entries allow invoking the details view or raise their palette so can not be erased via the Sugar Journal UI. The three corrupt entries will be an entry for Browse where I visited ASLO to download telescope and then quit Browse; the downloaded Telescope-12 bundle; and the instance entry for the Telescope that was running when the XO locked up.

Speculation.  The filesystem was not in a consistent state with respect
to the journal entries that were recently created.  Sounds like #11160.
We saw this with our previous attempt to use ext4.  See also #11220.
#9612 also has some interesting notes, especially the observation of
files that exist but have nothing in them.

Indeed, as you suggest, Sugar Journal might benefit from being trained
how to cope with this corruption pattern.  The copy you have of the
corrupted datastore would be useful for whomever takes this task on.

--

If you can reliably reproduce the Telescope-12 hang of the system, then
attach the serial port to a terminal emulator with logging.  I suggest
"screen -L /dev/ttyUSB0 115200" on another system.  The resulting output
will be very useful.

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