david at kineticode
Feb 16, 2009, 8:24 PM
On Feb 5, 2009, at 10:18 AM, David E. Wheeler wrote:
>> I have 2 theories - 1) the session data regarding versions got
>> borked (or was cached?) and failed to be aware of the fact that it
>> had v6, checking it in as something else, or 2) The check in hadn't
>> completed before the checkout occurred, resulting in corruption.
> Yeah, good working hypotheses. If the session had the story with
> version 4 instead of 6, then it would try to check it in as version
> 5. That'd be my first guess. The "checkin hadn't completed" bit
> shouldn't be the case, thanks to transactions, but it's possible
> that there could be a couple of transactions on checkin. Still, the
> bit that determines the new version number should not be confused. I
> think the session issue is much more likely the culprit.
Well, I did find a way in which version numbers can get mucked up, but
it'd be kind of hard for it to happen. It would basically require very
slow publishes. I'm not completely convinced that my recently-
committed fixes will address the issue, though I certainly hope so.
The attached patch has the relevant fixes, including for Scott's Bug
#1417. Please do apply this to your Bricolage instance, Matt, and let
me know if the problem of story version number corruption goes away or
President, Kineticode, Inc.
Kineticode. Setting knowledge in motion.[sm]