rolfm at denison
Feb 4, 2009, 10:09 AM
On Feb 4, 2009, at 10:30 AM, David E. Wheeler wrote:
> On Feb 3, 2009, at 11:47 AM, Matthew Rolf wrote:
>> Now, this story was republished by the system, before the error
>> occured, but it was checked in when that happened.
> NO, it wasn't:
>> Story Checked Out 2008-04-07 20:39:17
>> Story Moved into Workflow 2008-04-07 20:39:16 Workflow: Recreation
>> Story Republished Maintainance, Bricolage 2008-01-11 14:10:27
What? It was republished in january, and checked out in April.
> It's clearly checked out when the maintenance republish happened.
> Furthermore, it was expired before it was checked in!
>> Story Expired Maintainance, Bricolage 2007-11-11 15:41:36
>> Story Republished Maintainance, Bricolage 2007-11-11 15:41:29
>> Story Changes Saved 2007-11-11 15:41:25
>> Story Removed from Workflow 2007-11-11 15:41:25
>> Story Checked In 2007-11-11 15:41:25 Version: 4
That's not the way I read it. I read that that it was checked in, and
then republished and expired at the same time. This log is in
chronological order from the bottom up - just like the log screen
displays. My apologies if that caused confusion.
>> Your pasting of the log for the story only goes to version 4.
No, it doesn't.
>> But, I ask you, what user is capable of checking a story in and,
>> within two seconds, finding it on a desk and checking it out again?
They can check it in on a desk and check it out very rapidly. The
Check In/Out button changes with the page refresh. It is entirely
>> Or, better, how likely is it that some other user checked it out
>> exactly two seconds after it was checked in? Seems kind of weird,
>> though not out of the question.
The log shows it being checked out by the same user. I omitted their
name, because I didn't want to past it on the list.
> When you check in a story, it checks in the instance you've been
> using. All it does is set checked_out to false and increment the
> version number. Somehow, though, you've got two checked-out versions
> at the same time, 5 and 6, which is just wrong. Also, where did that
> crazy, out-of-sequence id come from, 102717? That doesn't look right
> at all. There's some code somewhere that's doing something funky. I
> sure would like to know where it is. :-(
Right, so what I'm saying is that looking at the log, the first check
in from 5 to 6 looks like it worked. But when the user checked it out
and in again, it went to version 5, and didn't increment, and didn't
create a new db entry. Then when they came back months later and
checked it out, it stayed at version 5, which is the out of sequence
id. They then could not check it in, because there was already a
version 6, and the unique key index wouldn't let them duplicate it.