On 02/14/2011 03:52 PM, Sascha Silbe wrote:
Excerpts from Simon Schampijer's message of Mon Feb 14 21:01:18 +0100 2011:
So if the timestamp is missing we will create the entry with the current
timestamp when we update_the_index. I guess there is no better timestamp
we can set.
Exactly. I considered setting it to 0 instead, but the current time
seemed to be a marginally better fit.
Do we have other properties we should care about?
None that the data store itself requires in order to work properly
(except for those we already check for).
When restoring the other properties we set as well "update_metadata =
True", I think you have to do that here as well.
Yeah, good idea. creation_time might have been written before timestamp
(and the latter not landed on disk). Actually that gives me an idea:
We could try checking creation_time and ctime. I'm not sure the rare
case of timestamp missing is warranting the additional code complexity
of checking those two properties, though.
Hmm, where does the ctime property actually comes from, is it a property
of the old DS?
I think just doing the check as you have it and setting
"update_metadata" to true should be enough.
Regards,
Simon
_______________________________________________
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel
