David,
David Jencks wrote:
On Jan 22, 2006, at 6:59 PM, Randy Watler (JIRA) wrote:
[
http://issues.apache.org/jira/browse/JS2-473?page=comments#action_12363599
]
Randy Watler commented on JS2-473:
----------------------------------
Yes, this is a bug as you read in the comment... constraints and
permissions checks on individual fragments are relatively new. We
made that change too close to the 2.0 cutoff, so we were not in the
mood to change the PageManager API then... :-).
I hope I am not demonstrating "fools rush in where angels fear to
tread" :-)
Nah... it is time to get this stuff fixed...
I am fine with the proposed API change, but I cannot read the diffs
very well. I am concerned about the DB version of the FragmentImpl
since it appears you moved functionality into the constructor of this
persistent class. I think we should probably go ahead and apply the
patch when others get s few moments to comment. Then, I can more
propery review the changes.
In general I am concerned about the thread safety of a lot of the code
I've looked at. In the DB FragmentImpl a list was being created
lazily on access, with no synchronization. I moved list construction
into the constructor, rather than write a new method to lazily create
the list. It's still not thread safe for modifications, but at least
everyone will get the same copy of the list. If all these classes are
only accessed by a single thread, that fact would be useful to
document somewhere easy to find.
Thread safety is an issue in the DB PM code. It is very new and somewhat
immature. I am going to add a separate JIRA issue so that it does not
get forgotten. Thanks for pointing it out!
Randy
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
