Blank/corrupt entries at end of Journal

[garycmartin]

Testing Sugar 0.87.7 in an F11 VM. Looking through existing entries in Journal shows a number of 'invisible' slots at the end of the Journal that I think aligns with broken/empty Journal items (that can be left hanging around after Sugar crashes). Mousing over the blank area left causing partial widget redraws for the missing items (see attached screen shot). Either Journal should list broken entries as broken objects (allowing the user to know there was some breakage and delete them manually); or the datastore should be cleaning up and/or not be returning empty/broken data to the Journal to incorrectly display.

Found this shell.log traceback which may be related:

Traceback:
File ".../install/lib/python2.6/site-packages/jarabe/journal/listview.py", line 425, in update_dates
path = tree_model.get_path(next_iter)
TypeError: iter should be a GtkTreeIter

On later examining of the Journal these blank items are now not showing anymore, so perhaps there is indeed some background clean-up of broken entries going on.

[sascha_silbe]

Might be another occurance of #1407? Do you still have my patch applied?
In any case, a full copy of shell.log and datastore.log (again heads up for #1807) would be quite useful.
We should make Sugar more robust in cases of metadata corruption, but I'd also like to find out what caused the corruptions.

[alsroot]

Would be useful to have copy of ~/.sugar/default/datastore to reproduce this issue.

[garycmartin]

Replying to sascha_silbe:

Might be another occurance of #1407? Do you still have my patch applied?

No,it was wiped when I moved to 0.87.7

In any case, a full copy of shell.log and datastore.log (again heads up for #1807) would be quite useful.

Sorry not available, both over written when I rebooted trying to test the bug when it was first encountered.

We should make Sugar more robust in cases of metadata corruption, but I'd also like to find out what caused the corruptions.

+1. I think I first saw the blank entries after the VM collapsed a few times while trying to use several instances of sugar at once (for collaboration testing). This is likely similar to a hard power off for real hardware, or a USB stick being pulled before synch.

[garycmartin]

Replying to alsroot:

Would be useful to have copy of ~/.sugar/default/datastore to reproduce this issue.

Sure I know, but too late now. We'll likely have to manually generate broken entries to test the robustness of various parts of the datastore/journal code.

[alsroot]

could be partially covered by #1826 issue

[martin.langhoff]

Also likely a symptom of the problem addressed in ​http://lists.sugarlabs.org/archive/sugar-devel/2012-September/039729.html

[dnarvaez]

Old, unreproducable, likely fixed.