[adrian.brock]

I'll fix it properly, but it is slightly complicated in that step (2) isn't the
correct thing to do if the class doesn't match the parent filter.

Actually, I think there's an issue with doing that, in that the parent cache could
get flushed due to a new deployment in that domain while the child blacklist is not.

I need to think about the way this works to make it more efficient.

But I don't think the blacklist optimaztion is as important as the caching so I'll
fix the use in BaseClassLoaderDomain.loadClass() and do a release of 2.0.6

I'll have to come up with a more optimal solution for 2.0.7 :-)

[alesj]

But I don't think the blacklist optimazation is as important as the caching so I'll
fix the use in BaseClassLoaderDomain.loadClass() and do a release of 2.0.6

Why the hurry?

I think Brian is the only one affected by this,
and I'm sure he's happy to work on latest snapshot. ;-)

I need to fix up JIRA anyway, so take your time. :-)
And I'll do a proper release once you're done.

[adrian.brock]

"alesj" wrote:

I need to fix up JIRA anyway, so take your time. :-)
And I'll do a proper release once you're done.

I'm done for now. The proper fix will be a bit more complicated so I need
to let my brain think about it as a background task for a while. ;-)

[adrian.brock]

The proper fix will be a bit more complicated ...

By more complicated, I mean I'll have to rework the way it works since the
BaseClassLoaderDomain doesn't know anything about the parent/child relationship.

It will also probably require flushing child caches when a parent is flushed
which would mean introducing some kind of listener mechanism.

[brian.stansberry]

Thanks, guys. :)

"alesj" wrote:

Why the hurry?

I think Brian is the only one affected by this,
and I'm sure he's happy to work on latest snapshot. ;-)

I certainly don't mind. Dominik's doing the perf testing, and he seemed fine with a snapshot as well. But I believe the component freeze for 5.1.0.GA is Monday.