Hi Jukka,
On Jul 17, 2009, at 5:26 PM, Jukka Zitting wrote:
Hi,
2009/7/17 Grégory Joseph <[email protected]>:
On Jul 15, 2009, at 12:15 PM, Jukka Zitting wrote:
Well, we'd need someone to submit a patch for that... I don't know
of
anyone actively working on that issue.
Sounds reasonable, of course. I *would* happily provide a patch, if
I had
any idea/suggestion what direction to go to fix this.
Take a look at the BindableRepository class and see if you could make
the shutdown() method remove the troublesome cache entry in
BindableRepositoryFactory.
This seemed reasonable at first, and I have a patch that actually does
something like that now. (from RegistryHelper, since that's where BR/
BRF are used)
But looking at the related JCR-411, I am now under the impression that
there's something more broken than it appears (or perhaps something
lacking a little inline comment here and there). In JCR-411, I see "On
the next lookup, BindableRepositoryFactory.getObjectInstance is
invoked". Reading this, I assumed that the javax.naming.Reference
instance that's used tells jndi to use BRF "on the next lookup" (the
Reference does point to BRF's classname; why else?). Well, as far as I
can tell, that's not the case. So now, I'm left wondering what the
jx.n.Reference is there for at all and why BRF has an instance cache
(which after all somehow is redundant with jndi)
The reason we use this is probably legacy related, and I'll look into
this, because it seems unnecessarily complex for us; otoh, if there's
any insight on what *does* happen with this code, if I'm missing the
point or if something is indeed broken, I'd be willing to help further.
Cheers,
-g
